IndustryDocuments first
Document work is the value pool
Diligence, filings, disclosure and precedent are where analyst hours concentrate, and they are exactly the shape retrieval with citation handles well.
Program 02 · Domain specialisation
SCIKIQ Certified Data and AI EngineerInvestment Banking
Every other industry treats access control as a security requirement. In investment banking it is the product: the wall between the deal side and the public side is what lets the firm exist in both. An agent that can retrieve across it is not a bug to be patched — it is a reportable event. This specialisation is built around that constraint from the first module.
- 80 hrstaught
- 40hands-on AI labs
- 8assessment tests
- +20 hrs2 E2E capstones
Information barrierThe constraint every design in this sector starts from
Document workWhere analyst hours concentrate, and where retrieval lands first
SurveillanceCommunications monitoring at a volume only automation can read
Post-tradeBreaks and confirmations that are a reconciliation problem
01
Part 1 · The business
The value chain, the people, the trends
The bank is deliberately several businesses that must not share what they know. Part 1 maps the chain and names the people on each side of the wall.
The investment banking value chain
Six stages from coverage to settlement. Select one.
01
Originate & cover
Client coverage, pipeline development, pitch preparation and mandate win.
Decisions made here
- Who do we cover?
- What do we pitch?
- Is this mandate winnable?
Where it breaks today
- Pitch material rebuilt from scratch for every meeting
- Precedent transactions known only to whoever worked on them
- Client intelligence scattered across coverage teams
What is measured
- Pipeline conversion
- Wallet share
- Pitches per mandate
Where AI landsPrecedent transaction retrieval, pitch drafting and client intelligence assembly.
Pipeline conversionWallet sharePitches per mandate
02
Execute the transaction
Diligence, valuation, documentation, marketing and closing across M&A, ECM and DCM.
Decisions made here
- What is it worth?
- What does diligence show?
- Is the disclosure complete?
Where it breaks today
- Diligence rooms read page by page under deadline
- Disclosure checked manually against a long requirement list
- Model and document versions drifting apart in the final week
What is measured
- Time to close
- Diligence cycle time
- Deal margin
Where AI landsDiligence document review, disclosure completeness checking and comparables assembly.
Time to closeDiligence cycle timeDeal margin
03
Research & publish
Coverage initiation, estimates, notes and the independence rules around them.
Decisions made here
- What do we say?
- What changed?
- Is this publishable?
Where it breaks today
- Filings and transcripts read in full to find the one changed line
- Estimate changes reconstructed manually each quarter
- Independence checks applied late in the publication path
What is measured
- Notes published
- Estimate accuracy
- Time to publish
Where AI landsFiling and transcript change detection, estimate assembly and note drafting support.
Notes publishedEstimate accuracyTime to publish
04
Sell & trade
Client sales, market making, order handling, execution and best execution evidence.
Decisions made here
- Where is the axe?
- How do we execute this?
- Was this execution good?
Where it breaks today
- Client interest inferred from memory rather than from the record
- Best execution evidenced after the fact, laboriously
- Market colour that never reaches the people who need it
What is measured
- Client revenue
- Execution quality
- Hit rate
Where AI landsClient interest matching, execution quality explanation and market colour summarisation.
Client revenueExecution qualityHit rate
05
Control & supervise
Information barriers, conflicts, surveillance, suitability and supervisory review.
Decisions made here
- Is this crossing the wall?
- Is this a conflict?
- Does this communication need review?
Where it breaks today
- Surveillance alerts dominated by false positives
- Conflict checks dependent on someone remembering to run them
- Communications volume far beyond what reviewers can read
What is measured
- Alert false-positive rate
- Review coverage
- Escalation cycle time
Where AI landsSurveillance triage, conflict detection and supervisory review prioritisation.
Alert false-positive rateReview coverageEscalation cycle time
06
Confirm & settle
Confirmations, allocations, clearing, settlement and break resolution.
Decisions made here
- Does this match?
- Why did it break?
- Who bears the cost?
Where it breaks today
- Breaks investigated by hand across counterparty formats
- Confirmation mismatches found late in the cycle
- The same break causes recurring quarter after quarter
What is measured
- Break rate
- Settlement fails
- Resolution time
Where AI landsBreak diagnosis, confirmation matching and recurring-cause identification.
Break rateSettlement failsResolution time
Who you are building for
Two of these people are legally not allowed to know what the other one knows. Select one to light the stages they own.
What the sector is seeing
A document-heavy business with a hard confidentiality boundary running through the middle.
IndustryEntitlement by design
The barrier is the architecture
Firms adopting AI here are designing entitlement into retrieval from the start, because retrofitting a wall into an index that ignored one is not possible.
IndustryTriage, not more reviewers
Surveillance volumes outgrew human review
Communications monitoring generates more alerts than reviewers can read, so triage that reduces false positives is the difference between coverage and theatre.
IndustryRecurring causes
Post-trade breaks are a reconciliation problem
Settlement breaks recur with the same causes across quarters, which makes them a data and pattern problem rather than a staffing one.
IndustrySeparate, not shared
Research independence constrains the design
What research may see and say is regulated separately from the rest of the firm, so a single shared corpus is a compliance problem however convenient it looks.
IndustryAuditable by default
Evidence is the deliverable
In a supervised business the output that matters is not the answer but the record of how the answer was reached and who could see what.
02
Part 2 · Data and meaning
Domains, ontology, taxonomy and governed terms
The unusual thing about this architecture is that the access control is not a layer bolted on at the end. Entitlement determines what may be indexed together at all, which means it shapes the retrieval design before a single document is loaded.
The data landscape
Six families, two of which must be provably unable to reach each other.
Deal & pipeline
Mandates, pipeline, diligence rooms and transaction documents.
Deal CRM, virtual data roomsMarket & reference
Prices, instruments, issuers, corporate actions and identifiers.
Market data vendors, security masterFilings & disclosure
Regulatory filings, prospectuses, transcripts and announcements.
Regulator archives, filing feedsResearch
Notes, models, estimates and the publication and independence record.
Research management systemsOrder & execution
Orders, executions, allocations and best-execution evidence.
OMS, EMS, execution venuesCommunications
Email, chat, voice and the surveillance record over all of it.
Archives, surveillance platformsThe semantic layer
In most industries the ontology exists so an agent can join data. Here it also exists so an agent knows what it is not allowed to join. Every entity below carries a side, and the graph is as much an entitlement model as a knowledge model.
Data domains
Six subject areas, each with the entities it holds, its critical data elements and the function accountable for it.
Client & mandate
Who the firm acts for, and on what.
ClientMandateDealPitch
- Critical data elements
- Legal entity identifier; Mandate type; Deal side
- Owned by
- Coverage
Issuer & instrument
The securities and the entities that issue them.
IssuerInstrumentCorporate actionRestricted entry
- Critical data elements
- ISIN; Ticker; Issuer identifier
- Owned by
- Reference data
Disclosure
What has been said publicly, and when.
FilingProspectusAnnouncementTranscript
- Critical data elements
- Filing date; Document type; Reporting period
- Owned by
- Research operations
Order & execution
Client orders and how they were filled.
OrderExecutionAllocationVenue
- Critical data elements
- Order timestamp; Executed price; Venue identifier
- Owned by
- Markets operations
Control
The barriers, lists and reviews that govern who may know what.
BarrierRestricted listConflictAlert
- Critical data elements
- Side classification; Restriction reason; Review outcome
- Owned by
- Compliance
Post-trade
Confirmation, settlement and the breaks between them.
ConfirmationSettlementBreakCounterparty
- Critical data elements
- Settlement date; Break reason; Counterparty identifier
- Owned by
- Operations
Ontology & knowledge graph
The entities and the typed relationships between them — what the agent traverses instead of guessing joins. Select any entity.
Core entityEventReference / classification
Core entity
Client
The party the firm acts for.
- Client grants Mandate
- Surveillance alert raised against Client
Core entity
Mandate
An engagement to act, and private side by nature.
- Mandate executes Deal
- Mandate triggers Restricted list
- Client grants Mandate
Core entity
Deal
The transaction a mandate is executing.
- Deal involves Issuer
- Mandate executes Deal
Core entity
Issuer
The entity whose securities are traded.
- Issuer issues Instrument
- Issuer publishes Filing
- Deal involves Issuer
- Research note covers Issuer
Reference / classification
Instrument
A tradable security of an issuer.
- Issuer issues Instrument
- Order references Instrument
- Restricted list restricts Instrument
Event
Filing
A public disclosure made by an issuer.
- Issuer publishes Filing
- Research note cites Filing
Event
Research note
Published analysis, subject to independence rules.
- Research note covers Issuer
- Research note cites Filing
Event
Order
A client instruction to trade.
- Order references Instrument
- Order filled by Execution
Event
Execution
A fill against an order.
- Execution may cause Break
- Order filled by Execution
Reference / classification
Restricted list
Securities on which activity is limited.
- Restricted list restricts Instrument
- Mandate triggers Restricted list
Event
Surveillance alert
A communication flagged for review.
- Surveillance alert raised against Client
Event
Break
A mismatch between two records of one trade.
- Execution may cause Break
Taxonomies
The classification hierarchies that make records comparable across systems.
Side classification
- Side
- Basis
- Scope
Private → Live mandate → Deal team only
Product hierarchy
- Division
- Product
- Sub-product
Advisory → M&A → Sell-side
Break cause
- Class
- Cause
- Owner
Economic → Price mismatch → Counterparty
Governed business terms
The definitions an agent must use rather than invent. Most wrong answers in this industry are a term used loosely.
- Private side vs public side
- Private side holds material non-public information; public side must operate without it. The distinction is legal, not organisational convenience.
- MNPI
- Material non-public information. Knowing a mandate exists is usually itself MNPI, before any deal detail is involved.
- Restricted list
- Securities on which firm activity is limited because of what the firm knows. Being on it is itself sensitive.
- Wall crossing
- Deliberately bringing a public-side person over the barrier, under a recorded, controlled process.
- Best execution
- The obligation to obtain the best available result for a client order, evidenced rather than asserted.
- Precedent transaction
- A comparable past deal used to inform valuation or structure, drawn from public record only.
- Break
- A mismatch between two records of the same trade that must be resolved before settlement.
- Research independence
- The separate regime governing what research may see and say, distinct from the general barrier.
03
Part 2 · The architecture
How it is actually built
The layers a deployment needs, what each one holds, and what the engineer owns there.
The reference architecture
Seven layers, with the information barrier expressed in the index rather than in policy.
01/07
Deal, market & communications sources
Private-side deal material, public market and filing data, and the communications archive.
Takes in
Nothing. This is the upstream edge.
Hands on
Side-labelled, entitlement-bearing feeds — public and private material never in one stream.
SystemWhat it doesKey business entities
Deal CRM and pipeline
Mandates, pipeline and coverage activity. Private side by definition: knowing a deal exists is itself material non-public information.
MandatePipelineClient
Virtual data rooms
The diligence material for a live transaction, supplied by the target and access-controlled per deal. The most sensitive corpus in the firm.
Diligence documentContractDisclosure
Market and reference data
Prices, instruments, issuers and corporate actions, under vendor licensing that limits redistribution and derived use.
InstrumentIssuerPriceCorporate action
Regulatory filing archives
Prospectuses, annual reports, announcements and transcripts. Public, structured by section, and the corpus most research work is grounded on.
FilingSectionAnnouncement
Order and execution systems
Orders, executions, allocations and the evidence behind a best-execution assertion.
OrderExecutionAllocation
Communications archive
Email, chat and voice under supervisory retention, plus the surveillance alerts raised over them.
CommunicationAlertReview
What the engineer owns here
Classifying every source as public or private side before it is ingested anywhere.
Where this layer fails
Loading a source before it has been classified public or private side. A document that enters unlabelled contaminates the partition it landed in, and the only remedy is to rebuild the index.
02/07
Ingestion & classification
Ingestion that carries side, deal and entitlement labels with every document.
Takes in
Read-mostly access paths, change feeds and document streams from the systems of record.
Hands on
Replayable, resolved, quality-checked records with their lineage back to the source row.
SystemWhat it doesKey business entities
Change data capture
Reads the source’s own change log rather than polling it, so the platform sees every state a record passed through instead of only where it ended up.
Change eventOperation typeCommit timestamp
Landing and raw zone
Every extract kept as it arrived, immutable and timestamped. The thing you replay from when a downstream definition turns out to have been wrong.
Raw extractBatchWatermark
Document processing
Layout-aware extraction over the unstructured half of the estate: sectioning, tables, signatures, and the page reference every later citation depends on.
DocumentSectionPage reference
Entity resolution
Decides that two records are the same real-world thing, with a survivorship rule and a confidence, so downstream joins are a decision rather than an assumption.
Match candidateSurvivorship ruleCluster
Data quality and contracts
Schema, freshness and volume expectations asserted at the boundary, so a bad load fails loudly here rather than quietly two layers later.
ContractExpectationViolation
Side classificationthis domain adds
Labels every document public or private side at the moment it is ingested, because a document that enters unlabelled can never be trusted afterwards.
DocumentSideDeal tag
What the engineer owns here
Labelling at ingestion, because a document that enters unlabelled can never be trusted again.
Where this layer fails
Ingesting first and classifying later. The partition is only as trustworthy as its worst unlabelled document, and you cannot find that document afterwards.
03/07
Entitlement & semantics
The entitlement model, the issuer and instrument graph, and governed definitions.
Takes in
Replayable, resolved records with lineage from ingestion.
Hands on
Typed, classified, defined data an agent can be grounded on without inferring meaning.
SystemWhat it doesKey business entities
Business glossary
One agreed definition per term, owned by a named person, so the number an agent quotes means what the business means by it.
TermDefinitionOwnerSteward
Ontology and knowledge graph
The typed entities and relationships the domain actually has, so an agent can traverse "which customers are exposed to this" rather than guess from adjacent text.
EntityRelationshipNodeEdge
Semantic and metrics layer
Metrics defined once, in one place, with their filters and grain. Removes the class of error where the agent computed something plausible and wrong.
MetricDimensionGrainFilter
Classification and policy
Sensitivity labels and the purpose each classification permits, applied at the field level and inherited by everything downstream.
ClassificationPurposePolicy
Lineage and catalogue
Where a value came from and what it touched on the way. The layer that makes an answer defensible rather than merely correct.
Lineage edgeDatasetColumn
What the engineer owns here
One entitlement model that both the index and the application obey, with no second copy.
Where this layer fails
Letting the model infer meaning from column names. It will, it will be plausible, and nobody will notice until the number reaches a regulator or a board pack.
04/07
Retrieval & grounding
Side-partitioned retrieval over filings, precedent, research and diligence material.
Takes in
Typed, classified, defined data from the governance layer.
Hands on
Ranked, permission-trimmed, citable evidence scoped to the caller and the moment.
SystemWhat it doesKey business entities
Hybrid index
Vector and lexical retrieval together, because exact identifiers, codes and clause numbers are the thing semantic search is worst at.
ChunkEmbeddingTermScore
Chunking and metadata
Splits on the document’s own structure and carries effective dates, version and source into every chunk, so a retrieved passage knows when it was true.
ChunkEffective dateVersionSource reference
Security trimming
Applies the caller’s permissions inside the query rather than filtering results afterwards, so the model never sees what the user may not.
PrincipalEntitlementFilter
Graph retrieval
Traverses the knowledge graph for questions that are joins rather than similarity — exposure, genealogy, ownership, causation.
PathHopTraversal
Reranking and citation
Reorders candidates on relevance and returns the passage identifier behind every sentence, so the answer can be checked rather than trusted.
CandidateRankCitation
Side-partitioned indexesthis domain adds
Physically separate indexes per side, with entitlement applied inside the query, so private material is not merely filtered but unreachable.
PartitionEntitlementPrincipal
What the engineer owns here
Enforcing the barrier inside the query, so private material is not retrievable at all publicly.
Where this layer fails
One index with an entitlement filter on top. A filter is a line of code somebody can bypass; a partition is not.
05/07
Agent & orchestration
Agents for diligence review, filing change detection, surveillance triage and break diagnosis.
Takes in
Ranked, permission-trimmed, citable evidence from retrieval.
Hands on
Actions taken or proposed, each with the evidence, the identity and the trace behind it.
SystemWhat it doesKey business entities
Agent runtime
Plans, calls tools and holds the loop. Where the step budget, the timeout and the stopping condition are enforced rather than hoped for.
TaskPlanStepTrace
Tool and MCP servers
Typed, permissioned tools with declared schemas. An agent’s real capability surface is this list, which is why the list is a security artefact.
ToolSchemaScopeInvocation
State and memory
Durable task state with checkpoints, so a run that dies mid-way resumes instead of restarting and re-doing side effects.
CheckpointSessionMemory record
Human-in-the-loop gates
Approval steps on the actions that need one, carrying enough context for the approver to actually decide rather than rubber-stamp.
ApprovalThresholdDecision
Workflow integration
Writes back into the systems people already work in, under a service identity with its own audit trail.
Work itemWrite-backActor
What the engineer owns here
Ensuring an agent inherits the user’s entitlements and never exceeds them.
Where this layer fails
An agent that acts under a service account rather than on behalf of the user. It will eventually do something the requesting user had no right to do, and the log will not show it.
06/07
Barrier, conflicts & supervision
Information barrier enforcement, conflict checks, restricted lists and the audit record.
Takes in
Proposed actions and drafted responses from the agent layer.
Hands on
Permitted, grounded, logged output — or a refusal with a stated reason.
SystemWhat it doesKey business entities
Policy enforcement
The rules that decide whether an action is permitted at all, evaluated before the action and independently of the model that proposed it.
PolicyDecisionObligation
Input and output screening
Injection detection on the way in, and on the way out the checks for leakage, unsupported claims and content the domain forbids.
PromptResponseViolation
Groundedness checking
Verifies each assertion resolves to retrieved evidence, and fails the response rather than shipping the sentence that does not.
AssertionSupportVerdict
Model risk management
Model inventory, intended use, validation evidence and the sign-off that lets a model be used for a purpose. Not optional in a regulated estate.
ModelIntended useValidationApproval
Audit log
Immutable record of what was asked, retrieved, decided and done — the artefact a reviewer reads when they do not take your word for it.
Audit entryActorEvidence reference
Barrier regression suitethis domain adds
The adversarial test that attempts cross-side retrieval and must fail, run on every build rather than on every audit.
Test caseVerdictBuild gate
What the engineer owns here
Being able to prove that no answer was ever assembled from material across the wall.
Where this layer fails
Proving the barrier once, at go-live. Entitlement drifts with every corpus change, so the proof has to be a build gate rather than a document.
07/07
Evaluation & observability
Retrieval accuracy, barrier regression tests, alert precision and cost per review.
Takes in
Permitted, grounded, logged output from the guardrail layer.
Hands on
Measured quality, cost and latency — and the evidence to change any of the three.
SystemWhat it doesKey business entities
Evaluation harness
Held-out sets and graded runs on every change, so a prompt edit is a measured change rather than a hopeful one.
Eval setCaseScoreBaseline
Regression gate
Blocks a release when quality drops, in CI, on the same evidence for everyone. The difference between a system and a demo.
ThresholdRunGate decision
Tracing
End-to-end spans across retrieval, model and tool calls, so a bad answer can be opened and read rather than argued about.
TraceSpanLatencyToken count
Cost and capacity
Cost per task and per tenant, against throughput and quota. The number that decides whether the pilot can become the rollout.
Cost per taskQuotaThroughput
Drift and feedback
Watches quality against production traffic rather than the test set, and routes real corrections back into the eval suite.
Drift signalFeedbackIncident
What the engineer owns here
The barrier regression suite — the test that must fail the build, not the report.
Where this layer fails
Evaluating once, before launch. Quality moves with the data, the model and the traffic, and a system with no live measurement has no idea which of the three moved.
Typical tech stack
Models
- Frontier models for document reasoning
- Small models for surveillance triage
- Long-context reading of filings
Grounding
- Side-partitioned document indexes
- Precedent transaction search
- Issuer and instrument graph
Agents
- Diligence review agents
- MCP over deal and market tools
- Surveillance triage
Data
- Classified ingestion with side labels
- Entity resolution across identifiers
- Communications archive integration
Assurance
- Barrier regression suite
- Retrieval precision evals
- Alert precision measurement
Runtime
- Deployment inside the firm boundary
- Per-side isolation
- Immutable audit logging
Constraints that shape the design
The information barrier is absolute
Private-side material must be unretrievable from the public side, enforced in the index and the query rather than by a policy the application is trusted to apply.
Everything is supervised
Communications and advice are subject to supervisory review and recordkeeping. Anything an agent produces enters that record and must be reconstructable.
Research independence is separate again
What research may see and say is regulated on its own terms. A convenient shared corpus across research and banking is a compliance finding waiting to happen.
Deadlines are external and immovable
Filing windows, closing dates and settlement cycles do not move for a system that is having a slow day.
04
Part 3 · Getting it done
Where the business and the technology meet
Every use case below is designed so that the entitlement question is answered before the retrieval question. In this sector that ordering is the whole discipline.
Use cases, mapped to the chain and the architecture
Filter by stage or by earned autonomy. Selecting a use case jumps the value chain and the architecture to the stage and layer it depends on.
Value chain stage
Autonomy earned
12 use cases
No use cases match that combination.
The delivery playbook
How an investment banking deployment actually runs.
01
Week 1 — settle the barrier before anything else
Agree the entitlement model and how it will be enforced in the index. Every later design decision is downstream of this, and it cannot be retrofitted.
02
Week 2 — classify sources by side, at ingestion
Label every source public or private before it is loaded. A document that enters unlabelled contaminates the partition it landed in.
03
Weeks 3-6 — start on the public side
Filings, precedent and market data carry no barrier risk and deliver value fast. They also build the citation discipline the private-side work will need.
04
Weeks 7-9 — move to deal material with the partition proved
Only once the barrier regression suite passes should private-side material be indexed. The suite runs on every build from that point on.
05
Weeks 10-12 — supervision, audit and handover
Prove that every answer is reconstructable and that no answer crossed the wall, then hand over runbook, regression suite and the audit evidence.
The curriculum
8 modules, 80 taught hours, 40 hands-on labs and 8 assessments — every lab provisioned and graded by the SCIKIQ Agentic AI Playground. Open a module to see its labs.
01The SCDAI delivery model in a supervised firmLanding in an organisation where the control function holds a genuine veto.10 hrs · 5 labs
Delivery mandateInformation barrier basicsValue sizing on analyst hoursStakeholder mapSupervisory expectationsThin-slice scoping
01
Decision selectionScore four banking decisions on value, cadence and barrier risk.
02
Value sizingSize a diligence cycle, a surveillance review and a settlement break.
03
Entitlement auditAssess whether entitlement data supports a retrieval use case.
04
Stakeholder mapMap coverage, execution, research, control and operations.
05
Thin sliceScope a public-side slice that carries no barrier risk.
Assessment 01 — Operating model & barrier scope
02Deal, market and communications dataSix data families, two of which must be provably unable to reach each other.10 hrs · 5 labs
Deal and data-room contentMarket and reference dataFilings and transcriptsOrder and execution dataCommunications archivesSide classification
01
Data room ingestionIngest a diligence room with side and deal labels attached.
02
Security masterResolve instruments and issuers across conflicting identifiers.
03
Filing ingestionIngest filings with section structure preserved for change detection.
04
Comms ingestionIngest a communications archive within its retention rules.
05
Side classificationClassify a mixed source set as public or private side and test it.
Assessment 02 — Banking data landscape
03Context engineering for financial languageInstruments, identifiers, figures and the precision a supervised answer demands.10 hrs · 5 labs
Identifier resolutionFigure extractionStructured outputsUncertaintyRefusal on entitlementContext budget
01
Identifier resolutionResolve an issuer across ISIN, ticker and legal name.
02
Figure extractionExtract reported figures with unit, period and basis intact.
03
Structured findingsEmit diligence findings as a typed, validated structure.
04
Entitlement refusalMake the agent refuse rather than partially answer across a barrier.
05
Context budgetCut context cost on a long-filing reading agent.
Assessment 03 — Context engineering
04Grounding on filings, precedent and diligenceThe document corpora that carry most of the analyst workload.10 hrs · 5 labs
Section-aware retrievalChange detectionPrecedent searchIssuer graph traversalCitationsRetrieval evals
01
Section retrievalRetrieve at section level across a long prospectus.
02
Change detectionDetect and classify changes between two filings.
03
Precedent searchFind comparable transactions with the comparison basis stated.
04
Citation contractCite document, section and date behind every extracted figure.
05
Retrieval evalsBuild an analyst-labelled question set and measure precision.
Assessment 04 — Retrieval & grounding
05Agent design behind a barrierEntitlement-aware agents, and what happens when a question spans both sides.10 hrs · 5 labs
Workflow vs agentEntitlement inheritancePartitioned toolsCross-side questionsSurveillance triageEscalation
01
Entitlement inheritanceMake the agent act strictly as the requesting user.
02
Partitioned toolsGive an agent tools whose reach is bounded by side.
03
Cross-side handlingRefuse and route a question that would require crossing the wall.
04
Triage agentRank surveillance alerts by likely reviewer value.
05
EscalationDefine when a finding must reach a control officer and prove it does.
Assessment 05 — Agent design & entitlement
06Integrating with deal, market and order systemsRead from systems that are themselves access controlled, write only into the record.10 hrs · 5 labs
MCP over deal and market toolsMarket data licensingOrder system readsWork-item write-backIdentityError handling
01
MCP over deal toolsExpose entitlement-aware deal lookups as typed read tools.
02
Licensed dataRespect market data licensing limits inside a retrieval design.
03
Order readsQuery execution data within its own permission model.
04
Write-backPost findings to a work item under a traceable identity.
05
IdentityBind agent access to desk, side and role entitlements.
Assessment 06 — Systems integration
07Barriers, conflicts and supervisionThe module the control function will read before anything is approved.10 hrs · 5 labs
Barrier enforcementConflict checksRestricted listsRecordkeepingOWASP LLM Top 10Incident response
01
Barrier designArchitect entitlement enforcement across index, query and context.
02
Barrier regressionBuild the suite that proves cross-side retrieval fails.
03
Restricted listEnforce restricted-list behaviour inside retrieval.
04
Injection defenceDefend against instructions embedded in a diligence document.
05
Red teamRun an OWASP LLM Top 10 pass over the diligence pipeline.
Assessment 07 — Barriers, conflicts & security
08Production, evidence and costIn a supervised business the audit record is part of the deliverable.10 hrs · 5 labs
Eval suitesBarrier regression gatingDriftAudit reconstructionCost per reviewHandover
01
Eval gateGate deployment on retrieval precision and barrier regression.
02
Audit reconstructionReconstruct how a given answer was produced, end to end.
03
Drift watchDetect drift after a filing format or corpus change.
04
Cost per reviewMeasure and reduce cost per surveillance review.
05
HandoverProduce runbook, barrier regression suite and audit evidence.
Assessment 08 — Production & evidence
Two capstones — 20 hours
End-to-end agent design, development, deployment and testing, 10 hours each. Reviewed by a Senior SCDAI Engineer against a published rubric.
Capstone A — Entitlement-aware diligence agent
Build an agent that reviews a diligence room and surfaces inconsistencies against the draft disclosure, citing document and section, while a barrier regression suite proves that no private-side material is retrievable from the public side under any query.
Capstone B — Surveillance triage agent
Build an agent that ranks communications surveillance alerts by likely reviewer value, with a measured false-positive reduction and an explicit, control-agreed threshold, plus evidence that no true positive class is systematically suppressed.
What you deliver, and how it is marked
- A running agent, deployed on the program's target stack, reachable by a reviewer.
- An architecture note: boundaries, tools, autonomy level and the trade-offs you took.
- A versioned eval set with a baseline and your final scores.
- A trace walkthrough of one success and one deliberate failure.
- Cost per task, measured — not estimated.
- A runbook and rollback path a client team could operate on Monday.
Problem fit15%Is this a decision worth automating, scoped to a defensible boundary?
Architecture20%Right pattern for the constraints; autonomy earned, not assumed.
Grounding & accuracy20%Measured groundedness against the eval set, with the failures named.
Safety & compliance20%Domain controls actually enforced, and demonstrably so.
Operability15%Traces, runbook, rollback, cost — could someone else run this?
Communication10%Can a client executive follow the reasoning in ten minutes?
One lab, start to finish
Every lab in this program follows the shape below. This is lab 07 in full — the brief you are given, the environment that is provisioned for you, the code you start from and the assertions that decide whether you passed.
Module 07 · Barriers, conflicts and supervision
Proving the information barrier holds
Make private-side material unreachable from the public side, and prove it under attack.
The brief
You are given a mixed corpus: public filings and a live data room, correctly side-labelled. Build the retrieval layer so a public-side user cannot reach private material by any route, then write the adversarial suite that tries. The suite includes queries that quote private text verbatim, that ask for it indirectly, and that attempt to infer it from what is absent.
Provisioned for you
- 420 public filings, 180 private data-room documents
- Side and deal labels on every document
- Two test identities: public-side and deal-team
- Adversarial query set (60 labelled)
You start from this
def retrieve(query, principal, index):
"""Return passages this principal is entitled to see.
Entitlement is a property of the index partition, not a filter applied
to results. If it is expressible as a post-filter, it is wrong.
"""
# 1. resolve the principal to the sides they may read
# 2. query ONLY those partitions -- never query then discard
# 3. return citations that can be checked against the partition
raise NotImplementedErrorWhat the grader asserts
- No private-side passage is returned to the public-side identity on any of the 60 adversarial queries
- Entitlement is enforced by partition selection, not by filtering results after retrieval
- A query quoting private text verbatim returns nothing rather than confirming the text exists
- The deal-team identity retrieves its own deal only, not other live deals
- Every returned passage cites a partition the principal is entitled to
05
Assessment
Assessed, not attended
Every module ends with a timed, randomised assessment delivered through the SCIKIQ Agentic AI Playground. The certificate requires a pass on all of them plus two reviewed capstones.
Assessment blueprint
What each test covers, how long it runs, and how many items are currently in the versioned bank behind it.
| # | Assessment & coverage | Items | Time | In bank |
|---|---|---|---|---|
| 01 | Operating model & barrier scopeDelivery mandateInformation barriersValue sizingThin slicing | 25 | 35 min | 3 |
| 02 | Banking data landscapeDeal dataMarket and referenceFilingsSide classification | 25 | 35 min | 2 |
| 03 | Context engineeringIdentifier resolutionFigure extractionStructured outputsRefusal | 25 | 35 min | 2 |
| 04 | Retrieval & groundingSection retrievalChange detectionPrecedentRetrieval metrics | 25 | 40 min | 2 |
| 05 | Agent design & entitlementEntitlement inheritancePartitioned toolsCross-side handlingTriage | 25 | 40 min | 2 |
| 06 | Systems integrationMCP designData licensingOrder dataIdentity | 25 | 35 min | 2 |
| 07 | Barriers, conflicts & securityBarrier enforcementConflictsRecordkeepingLLM security | 30 | 45 min | 2 |
| 08 | Production & evidenceEval gatingBarrier regressionAudit reconstructionEconomics | 30 | 45 min | 2 |
Single best answerOne defensible option among plausible distractors drawn from real field mistakes.
Scenario judgementA client situation with constraints; you pick the action a Senior SCDAI Engineer would take.
Architecture selectionGiven non-functional requirements, choose the pattern and justify the trade-off.
Failure diagnosisA trace, an eval report or a cost curve — identify the root cause.
- Each module ends with a timed assessment; the program certificate requires a pass on every one.
- Items are drawn at random from a versioned bank, so no two learners sit an identical paper.
- Pass mark is 70%. Two retakes are included, each with a fresh draw.
- Scenario items carry double weight — they are the ones that predict field performance.
- Capstones are reviewed against a published rubric by a Senior SCDAI Engineer, not auto-graded.
Try four questions from the bank
Real items, drawn from 17 in this program's bank — weighted toward the scenario and diagnosis types, because those are the ones that predict field performance. Instant feedback, nothing saved.
Four sample items — one attempt each, then the reasoning is shown.
Q1A bank asks for AI across advisory and research. What do you settle first?
Every later design decision is downstream of the barrier. It cannot be retrofitted into an index that was built without it, so it is a week-one conversation or a rebuild.
Q2When must a document be classified public or private side?
A document that enters unlabelled contaminates its partition, and you cannot reliably find it again. Classification is an ingestion-time gate, not a retrieval-time decision.
Q3An agent is asked a question that would require material from both sides of the barrier. What should it do?
A partial answer that does not name the boundary invites the user to infer what is missing. Refusing and routing is both safer and more useful.
Q4A change-detection agent flags a wording change in a risk factor. Should it classify it as material?
Materiality is a regulated judgement with consequences. The agent adds most value by finding candidates reliably, not by concluding.
06
Career & outcomes
What this makes you, and where it goes next
The same ladder whichever specialisation you enter through — what changes is the domain you go deep in. Below: how the program is delivered, the skills it moves, the roles it leads to, and the specialisations closest to this one.
Available two ways
The same labs, assessments and capstones, delivered to an enterprise cohort or to individual professionals.
B2B
Direct to enterprise
Cohorts of 20 to 2,000+ on your own tenancy, with your data patterns and your cloud. Skill-gap baselining up front, per-team mastery reporting throughout, and capstones scoped against your real backlog so the output is deployable work.
- Private cohort on your tenancy
- Baseline skills assessment before kick-off
- Capstones scoped to your backlog
- Per-team mastery dashboards for L&D
- Optional Senior SCDAI Engineer architecture reviews
B2C
Direct to individual professionals
The same labs, assessments and capstones for individual engineers and analysts, run on shared infrastructure with a fixed cohort calendar. You leave with a graded portfolio, not a certificate of attendance.
- Fixed-date cohorts, self-paced within them
- Same graded labs and item bank
- Portfolio-ready capstones you own
- Certificate on a pass in every module
- Community and office-hours access
Skill matrix
Find your row and aim one column right. The Playground scores you against this after every module.
| Skill | Beginner | Intermediate | Advanced |
|---|---|---|---|
| Banking domain fluency | Knows the product areas. | Maps the chain to who may know what. | Sizes analyst-hour and control value credibly. |
| Entitlement engineering | Aware of the barrier. | Enforces entitlement inside the query. | Runs a barrier regression suite as a release gate. |
| Document intelligence | Retrieves from a filing. | Detects and classifies change across filings. | Runs section-level retrieval with measured precision. |
| Context engineering | Writes clear prompts. | Structures retrieval, tools and state deliberately. | Designs context strategy for reliability and cost at scale. |
| Retrieval & grounding | Builds basic vector search. | Tunes chunking, hybrid search and reranking. | Designs graph + vector grounding with measured recall. |
| Agent orchestration | Runs a single tool-calling agent. | Builds supervised multi-step and multi-agent flows. | Designs autonomy boundaries and failure containment. |
| Tool & system integration | Calls a documented API. | Writes an MCP server over a system of record. | Designs a least-privilege tool estate across systems. |
| Evaluation | Eyeballs outputs. | Builds labelled eval sets and regression gates. | Runs online evals with drift and judge calibration. |
| Observability & cost | Reads logs. | Traces runs, tracks tokens and latency. | Owns cost per task and capacity planning in production. |
| Security & guardrails | Adds output filters. | Mitigates the OWASP LLM Top 10 in a build. | Threat-models an agent estate and proves controls. |
| Client delivery | Takes notes in a workshop. | Runs discovery and scopes a thin slice. | Owns the account technically, from scope to handover. |
Where it takes you
The SCDAI ladder is the same whichever specialisation you enter through — what changes is the domain you go deep in.
01
Data and AI EngineerEntry
Python, LLM APIs, first RAG build, prompt and context basics.
02
Associate Data and AI EngineerPractitioner
Client-facing discovery, tool integration, agent build under supervision.
03
SCIKIQ Certified Data and AI EngineerSpecialist
Owns a deployment end to end — scoping, build, evals, handover.
04
Senior Data and AI Engineer / Domain LeadSenior
Multi-agent architecture, regulated deployments, reference patterns for the practice.
05
AI Architect / Practice LeadLeadership
Portfolio of accounts, delivery standards, hiring and capability strategy.
Terms worth knowing
- Information barrier
- The enforced separation between private-side and public-side parts of a firm.
- Private side
- Where material non-public information is held, typically deal teams.
- Public side
- Research, sales and trading, which must operate without private-side information.
- MNPI
- Material non-public information — the thing the barrier exists to contain.
- Restricted list
- Securities on which firm activity is limited because of what the firm knows.
- Best execution
- The obligation to obtain the best available result for a client order, and to evidence it.
- Break
- A mismatch between two records of the same trade that must be resolved before settlement.
- Precedent transaction
- A comparable past deal used to inform valuation or structure.
- SCDAI
- SCIKIQ Certified Data and AI Engineer — the credential this program awards.
- SCDAI engineer
- An engineer embedded with the customer who turns their constraints into a working system.
- Thin slice
- The narrowest end-to-end path through a workflow that still produces business value.
- Agentic RAG
- Retrieval where the agent decides when, what and how often to retrieve.
- MCP
- Model Context Protocol — open standard connecting an agent to tools and data.
- A2A
- Agent2Agent — protocol for agent-to-agent delegation across vendors.
- Eval set
- A labelled, versioned set of cases an agent must pass before release.
- LLM-as-judge
- Using a model to score outputs, calibrated periodically against human ratings.
- Guardrail
- An enforced constraint on an input, output or action — not a prompt request.
- Human-in-the-loop
- A required human approval step before a consequential action executes.
- Cost per task
- Total token, tool and compute cost of one completed unit of work — the engineer's unit economic.
- Golden path
- The reference architecture a practice standardises on so deployments stay reviewable.
- Handover
- The point at which the client can operate, evaluate and change the system without the engineer.
Related programs
Skill a team, or join a cohort
B2B cohorts run on your tenancy with capstones scoped to your backlog. B2C cohorts run on a fixed calendar.
07
Enrolling
What you need, what you get, and what it costs
Stated plainly enough to rule yourself in or out without a sales call: the prerequisites, how the program runs, exactly what the credential is worth, and the questions everyone asks.
Before you start
Stated plainly so you can rule yourself in or out without a sales call. Nothing here is a formal qualification — it is what the first lab assumes you can already do.
Technical
You should already be able to do these
- Comfortable in Python: functions, typing, virtual environments, reading a stack trace
- Have called an LLM API and handled a structured response
- Git, the command line, and reading someone else’s code without panic
Domain
What we assume, and what we teach
- Know what a mandate, a filing and a trade are; product depth is taught
- Willingness to treat access control as the design rather than as a setting
Time and access
What the program asks of your week
- A machine that can run a browser and a terminal — labs run in the Playground, not locally
- Roughly 8 hours a week for the taught weeks, plus 20 hours for the two capstones
How it runs
Cohort dates and pricing are confirmed on enquiry rather than printed here, because both move with the intake.
- Format
- Self-paced within a fixed cohort window, delivered in the SCIKIQ Agentic AI Playground
- Taught hours
- 80 hours across 8 modules
- Capstones
- 20 hours — two end-to-end builds, reviewed not auto-graded
- Typical duration
- 10–12 weeks at around 8 hours a week
- Level
- Intermediate → Advanced
- Built for
- Engineers and analysts deploying AI across advisory, research and markets
- Next cohorts
- — confirmed on enquiry
- Pricing
- — confirmed on enquiry
The credential
The credential is awarded per specialisation, so it names the domain or stack you were assessed in rather than claiming general competence. On this program the badge reads SCDAI — Investment Banking.
How it is earned
- A pass at 70% or above on every module assessment — not an average across them
- Both capstones reviewed against the published rubric by a Senior SCDAI Engineer
- All labs submitted and machine-graded green
- Awarded as
- A verifiable digital badge with a public credential record
- Valid for
- Two years from award
- Recertification
- One current-year assessment plus a refreshed capstone
- Retakes
- Two included per assessment, each with a fresh item draw
- Named specialisation
- The badge states the domain or stack, e.g. SCDAI — Healthcare
- Verification
- Third parties confirm a holder from the credential record, no login required
A certificate that cannot be checked is decoration. Every award resolves to a record showing the specialisation, the award date and the assessments passed.
The certificate you leave with
This is the credential the program awards, shown exactly as it is issued — with the specialisation named, the assessment record attached and a verification link anyone can check without an account.
SCIKIQ Data Academy
Certificate of Professional Certification
This is to certify that
Your name
has been assessed and certified as
SCIKIQ Certified Data and AI Engineer
Investment Banking
- Specialisation
- Investment Banking
- Assessments passed
- 8 of 8, each at 70% or above
- Labs graded
- 40 of 40 machine-graded green
- Capstones reviewed
- 2, against the published rubric
- Taught hours
- 80 across 8 modules, plus 20 capstone hours
AwardedOn completion
Valid untilTwo years from award
Credential IDSCDAI-INVESTMENT-B-<issued>
Senior SCDAI Engineer, Certification Board
- 01The specialisation is on the certificate
Not “Data and AI Engineer” but the domain or stack you were actually assessed in. A general claim would be a weaker one.
- 02The assessment record travels with it
Modules passed, labs graded and both capstones reviewed — so the credential states what was measured rather than that you attended.
- 03Anyone can verify it, without a login
The credential ID resolves to a public record showing the specialisation, the award date and the assessments passed.
Share it on LinkedIn
Add it to your LinkedIn profile in one step. The link pre-fills the certification fields from the credential record, so the entry on your profile matches the record a reader can check.
Add to LinkedIn profile The button is live on your real certificate; here it opens LinkedIn pre-filled with this specialisation so you can see exactly what the profile entry will say.A certificate that cannot be checked is decoration. Every award resolves to a record showing the specialisation, the award date and the assessments passed.
Questions we get asked
The objections that come up in every conversation about this program, answered without the brochure voice.
Is this a course or a certification?
Both, and the second is the point. Eight timed assessments and two reviewed capstones stand between you and the credential, so a pass means someone measured the skill rather than recorded your attendance.
How is it self-paced without falling apart?
Every lab is provisioned, graded and unblocked by the Playground rather than by an instructor. That is what lets a cohort of 2,000 cost the same faculty time as a cohort of 20 — and why you are never waiting on someone to mark your work.
What if I fail an assessment?
Two retakes are included per assessment, each drawing a fresh item set from the bank, so a retake is a genuinely new paper rather than the same questions again.
Do I need my own cloud account?
For the domain specialisations, no — labs run in provisioned sandboxes. For the four tech-stack programs you will want access to that platform, since deploying into a real subscription is much of the point.
Can we run this privately for our team?
Yes. B2B cohorts run on your own tenancy with your data patterns, a skills baseline before kick-off, per-team mastery reporting, and capstones scoped against your actual backlog so the output is deployable work rather than an exercise.
Which specialisation should I take?
Take the domain you deploy into. If you move across industries, take a tech-stack program instead and pick up domain context on the engagement. The chooser on the programs page will narrow it.
How current is the material?
The trends, platform capabilities and regulatory positions are reviewed each quarter, and every external claim on these pages links to its source so you can check the date yourself.
What do I actually leave with?
A graded portfolio: forty machine-graded labs, two reviewed end-to-end agent builds with measured evaluation and cost per task, and a verifiable credential naming your specialisation.
Still deciding?
Tell us the systems you deploy into and we will say plainly whether this specialisation is the right one — or which of the 21 is.