Career learning, from first principles to BAU
Build fluency across markets, protocols, products, and AI—then practice the decisions, handoffs, reconciliations, and exceptions that make up the workday.
Each track starts with the market structure, then moves into the role-specific work.
Market plumbing, instruments, controls, and institutional workflows.
Protocols, on-chain settlement, liquidity, and smart-contract risk.
Payments, data rails, product controls, and regulated delivery.
Models, workflow automation, governance, and human-in-the-loop design.
A realistic foundation for understanding one of the market's less visible businesses.
Traditional finance · operations
Follow a loan across borrower demand, inventory, pricing, collateral, settlement, billing, recalls, and daily controls. Practice identifying breaks before they become exposure.
One operating foundation can open several different career directions.
Own the daily book, settlement, breaks, and escalation rhythm.
Support inventory, pricing, borrow demand, and lifecycle events.
Turn desk pain points into scalable workflows and controls.
Design data checks and exception-based human review.
Four field guides · 32 modules
Each guide connects domain knowledge to the work: who owns what, how a day unfolds, where judgment matters, what good output looks like, and how to prove readiness in an interview.
Track 01 · institutional operating model
Learn how trades become positions, cash and assets move, records stay aligned, and regulated institutions prove control. This path is especially useful for operations, product, business-analysis, and risk pivots.
Titles differ by firm. Compare the operating objective and outputs, not the label alone.
Confirms, settles, reconciles, and resolves lifecycle breaks before cutoffs.
Values exposure, agrees calls, checks eligibility, and escalates disputes.
Processes income, elections, tax, and client reporting around held assets.
Turns manual controls and desk pain points into governed platform changes.
The rhythm follows market deadlines. Seniority changes the scope, but the control loop is similar.
Participants, venues, custodians, clearing, settlement, and the difference between execution and finality.
Inventory, borrow demand, pricing, collateral, recalls, returns, and fee economics.
Capture, enrichment, confirmation, matching, settlement, maintenance, and close.
Exposure, marks, haircuts, eligibility, substitutions, margin calls, and disputes.
Dividends, interest, corporate actions, manufactured payments, and elections.
Completeness, matching keys, aging, root cause, tolerances, and evidence.
Preventive versus detective controls, escalation, maker-checker, and residual risk.
Translate positions, revenue, service, and incidents into decision-useful reporting.
Work exposure and deadline before inbox order or item count.
Trace the first incorrect event instead of repeatedly treating the symptom.
State impact, deadline, evidence, owner, requested action, and consequence.
Define population, frequency, threshold, reviewer, evidence, and failure path.
A same-day delivery is unmatched. Identify the record to compare, the deadline, and the party able to act.
Separate the temporary repair from the upstream change that prevents recurrence.
Write a five-line update covering impact, containment, cause, next action, and timing.
Create a process map with systems, owners, inputs, outputs, controls, exceptions, and one proposed improvement. Use invented labels, never employer-confidential data.
Track 02 · programmable financial systems
Learn to read protocol state, trace transactions, understand economic incentives, and operate safely around immutable execution, volatile collateral, governance, and smart-contract dependencies.
Monitors activity, treasury workflows, parameter changes, and incident queues.
Studies collateral, liquidity, oracle, concentration, and liquidation behavior.
Turns transactions and contract events into usable behavioral and risk signals.
Designs user flows and safeguards across wallets, contracts, and off-chain services.
Keys, signatures, addresses, multisig, hardware custody, approvals, and recovery risk.
Transactions, gas, blocks, confirmations, finality, reorgs, explorers, and event logs.
Automated market makers, pools, price impact, slippage, routing, and impermanent loss.
Supply, borrow, utilization, rates, collateral factors, health, and liquidation.
Backing models, redemption, peg mechanics, liquidity, reserves, and failure modes.
Supply, emissions, incentives, fee capture, governance rights, and reflexivity.
Data sources, update logic, manipulation risk, proposals, voting, and execution delays.
Smart-contract, liquidity, bridge, governance, custody, operational, and regulatory risk.
Use transaction hashes and events to reconstruct facts before assigning cause.
Ask how users, liquidators, liquidity providers, and governance may react.
Treat wallet permissions, signing context, and transaction simulation as controls.
Separate exploited code, market loss, user error, and third-party dependency failure.
Outline what you would check across prices, oracle freshness, collateral health, and liquidity.
Build a verification sequence that does not rely on a social message alone.
Write benefits, downside scenarios, monitoring signals, and a safe execution plan.
Choose a public transaction and produce an annotated trace: calls, events, assets, state change, fees, risks, and what you would monitor next.
Track 03 · financial products at software speed
Learn how customer actions become ledger entries and external money movement, with identity, fraud, partner, compliance, and reliability controls around every handoff.
Owns exceptions across authorization, clearing, settlement, returns, and disputes.
Connects support signals, incidents, launch readiness, and operating procedures.
Moves clients from requirements through integration, testing, launch, and stabilization.
Tunes detection and review while balancing loss, friction, and false positives.
Authorization, capture, clearing, settlement, reversals, retries, and routing.
Requests, responses, webhooks, idempotency, timeouts, authentication, and partners.
Accounts, entries, balances, pending versus posted, double entry, and reconciliation.
Identity evidence, verification, sanctions screening, risk tiers, and manual review.
Signals, rules, models, queues, false positives, account action, and loss feedback.
Reason codes, evidence, deadlines, provisional outcomes, customer communication.
Discovery, requirements, acceptance criteria, release readiness, runbooks, and adoption.
Approval rate, failure rate, loss, false positives, latency, SLA, and control evidence.
Follow one customer event through APIs, decisions, partners, ledger, and UI.
Know whether funds are available, pending, settled, reversed, or disputed.
Frame speed, conversion, fraud, compliance, cost, and support load together.
Define edge cases, monitoring, ownership, and rollback before launch.
Map the evidence needed to distinguish duplicate presentation, retry, or UI display issue.
Segment by issuer, processor response, geography, channel, and release timing before proposing a fix.
Draft detection, containment, customer state, reconciliation, and recovery steps.
Map a card payment or bank transfer with states, systems, partner messages, ledger entries, customer messages, controls, and three edge cases.
Track 04 · governed human-plus-model work
Learn how to select useful workflows, design evaluations, manage data and model risk, place human judgment, monitor production behavior, and earn adoption without overstating automation.
Selects use cases, defines evaluation, aligns controls, and drives adoption.
Monitors quality, drift, latency, incidents, versions, and review queues.
Combines rules, models, workflow tools, and human review into reliable processes.
Maintains inventory, assessments, approvals, controls, evidence, and issue remediation.
Training versus inference, classification, ranking, generation, confidence, and limitations.
Representative sets, labels, baselines, error classes, precision, recall, and human review.
Summarization, retrieval, drafting, triage, recommendation, action boundaries, and feedback.
Extraction, classification, confidence, source grounding, exceptions, and record linkage.
Validation, change control, bias, robustness, explainability, inventory, and issues.
Approval points, abstention, overrides, escalation, sampling, and accountability.
Use-case value, feasibility, risk, data readiness, build-buy, cost, and sequencing.
Workflow fit, training, quality, drift, latency, cost, incidents, and value realization.
Start with measurable friction and a clear decision, not a model looking for work.
Test the real task and costly failure modes, not just a generic benchmark.
Define when the system acts, recommends, abstains, or requires approval.
Connect monitoring and user overrides to model, prompt, data, or workflow change.
Define input, output, reviewer decision, evidence links, failure classes, and launch metrics.
Separate data shift, model change, prompt change, upstream defect, and user-behavior change.
Compare assistive, approval-based, and autonomous designs against impact and risk.
Write an AI use-case brief with baseline workflow, user, data, risk classification, evaluation set, thresholds, review design, rollout, monitoring, and fallback.
Section 01 · the business model
Securities lending is a temporary transfer of securities from a lender to a borrower, paired with collateral and a contractual obligation to return equivalent securities.
A lender monetizes otherwise idle inventory. A borrower gains temporary access to securities—for example, to support settlement, market-making, or a trading strategy. The lending agent may sit between them, handling allocation, pricing, settlement, collateral, income, and controls.
Provides the security while retaining the economic exposure defined by the agreement.
Posts collateral, pays the agreed economics, and returns equivalent securities.
Runs the operational and control layer on the lender's behalf.
The trade is not “done” when it is booked. It remains a living position: values change, collateral moves, rates can change, income events occur, and either side may initiate a return or recall.
POSITIONDo both sides agree on security, quantity, rate, and dates?BOOK & RECORDEXPOSUREDoes collateral sufficiently cover current loan value under the agreement?DAILY CONTROLMOVEMENTDid securities and collateral settle where and when expected?SETTLEMENTINCOMEWere fees, rebates, and manufactured payments calculated and paid correctly?BILLINGSection 02 · loan lifecycle
The lifecycle is a chain of connected handoffs. A clean book depends on each event being reflected consistently across the trading, operations, custody, and collateral records.
If a rate update appears in the trading record but not the billing record, the position can settle cleanly and still produce the wrong invoice. Controls must follow the whole lifecycle, not one system.
The borrower typically returns securities of the same issuer, issue, class, and quantity—not necessarily the exact same certificates or units. Legal terms and market convention determine the precise obligation.
Operational consequence: teams reconcile economic attributes and quantities, not a story about where a particular share traveled.
Section 03 · business as usual
Choose a window to see the operating objective, the decisions being made, and the signal an experienced analyst watches.
Review settlement status, collateral calls, recalls, returns, corporate-action notices, rate changes, and unmatched positions across internal and counterparty records.
Work failed or pending settlements, validate instructions, chase counterparties or custodians, and make sure recalled securities are on a credible path back.
Process new loans, returns, rate changes, reallocations, marks, and margin movements. Reconcile changes before downstream billing and reporting consume them.
Complete reconciliations, document aged exceptions, confirm escalations, and leave the next team a clear record of what remains open, why, and who owns the next action.
New trades, returns, recalls, rate changes, marks, and income events.
Movements or records that did not reach the expected state.
Positions where timing or value creates meaningful risk.
Market cutoffs, recalls, margin, billing and client reporting.
The named party with the next executable action.
Section 04 · economics lab
Change the assumptions to see how market value, fee, collateralization, and time flow into simple loan economics.
This is an illustrative practice calculation, not market data or a statement of any firm's billing convention. Real agreements can use different bases, rebates, currencies, haircuts, and fee treatments.
Four hypothetical open loans, calculated on a 360-day basis. Bar length starts at zero and represents daily fee revenue in US dollars.
Section 05 · exception management
Experienced operators classify an exception, quantify the risk, identify the true owner, and drive an action before the relevant deadline.
1 · EXPOSUREHow much risk exists if no action is taken?QUANTIFY2 · CUTOFFWhen does the next operational option disappear?TIME-BOX3 · CAUSEIs the break trade, settlement, static data, valuation, or timing?CLASSIFY4 · OWNERWhich team or counterparty can take the next action?ASSIGN5 · EVIDENCEWhat record proves the diagnosis and action?DOCUMENTA recalled position is due back today. Your internal book shows the return instructed; the custodian shows no matching receipt; the borrower says it sent the securities. What is the best first move?
A useful escalation gives the receiver a decision-ready package: position, value or exposure, contractual or market deadline, current state, verified cause, actions already taken, the next requested action, and the consequence of waiting.
Forwarding a long email chain transfers reading, not ownership. Summarize the issue in the first lines and attach the evidence underneath.
Section 06 · books and records
A reconciliation is not just a difference report. It is a control process that explains the difference, determines whether it is valid, and assigns resolution.
The rows below are a hypothetical training set. Identify the true break and the timing difference.
| Security | Internal qty | Counterparty qty | Difference | Likely status |
|---|---|---|---|---|
| Security A | 250,000 | 250,000 | 0 | Matched |
| Security B | 80,000 | 50,000 | 30,000 | Investigate partial return |
| Security C | 0 | 125,000 | −125,000 | Check booking cutoff |
Security B may be a genuine quantity break. Security C might be a timing difference if a return was booked internally after the counterparty produced its snapshot. Validate effective times before changing records.
All in-scope positions are included.
Fields and calculations match source records.
Review and escalation happen before risk-relevant deadlines.
Every unresolved item has a named next action.
The reviewer can retrace what was checked and why.
Section 07 · knowledge check
Answer three desk-level questions. Feedback appears immediately; your score only lasts for this session.
Career map
Compare the daily objective, core outputs, and proof points employers can evaluate. Titles vary by firm; the operating responsibilities are the more durable signal.
Keeps the live loan book accurate, settled, collateralized, and controlled.
Monitors exposure and drives margin movements under documented terms.
Keeps payment flows moving while controlling failure, fraud, and reconciliation risk.
Monitors on-chain activity, governance, liquidity, and operational risk.
Turns a financial workflow into a governed, measurable human-plus-model product.
Translates desk behavior, data, and control needs into implementable change.
Connects inventory, pricing, bookings, and lifecycle events around a live desk.
Evaluates processes, controls, incidents, issues, and residual exposure.
Coordinates liquidity, cash positioning, funding, and settlement readiness.
Studies collateral, liquidity, oracles, governance, and liquidation behavior.
Turns public contract activity into product, treasury, and risk insight.
Connects launches, incidents, support signals, controls, and operating readiness.
Balances loss prevention with approval, customer friction, and review capacity.
Runs the production control loop around model quality, reliability, and change.
Choose one. This is a directional prompt, not a personality test.
Desk dictionary
Plain-language definitions framed around what the term changes operationally.
The investor that owns the economic interest in assets made available for lending, often through an agent.
The economic charge associated with borrowing a security. On a desk, the agreed rate must flow correctly into accrual and billing records.
Assets delivered to secure an exposure. Operations monitors value, eligibility, movement, and release against contractual terms.
A movement that did not settle as expected. The useful question is why: instruction, inventory, timing, matching, or another cause.
A valuation adjustment used in determining how much collateral is required relative to an exposure.
A check or indication that a security may be available to borrow. The exact process and obligation depend on context and market rules.
A payment intended to pass through an economic entitlement when the borrowed security has an income event during the loan.
A lender-initiated request for borrowed securities to be returned. It creates a deadline-sensitive workflow.
A control comparing two records, explaining differences, and driving unresolved items to a documented outcome.
Code deployed to a blockchain that changes protocol state when defined conditions and transactions are processed.
A mechanism that supplies external or derived data—often prices—to a smart contract. Bad or stale inputs can change protocol outcomes.
Assets committed to a smart contract to facilitate protocol activity such as swapping or lending, subject to that protocol's rules and risks.
The authoritative record of balances and movements. In fintech, reconciliation often proves that product events and ledger entries agree.
The process of validating and preparing obligations for settlement, including calculation and matching steps.
A workflow in which a person reviews, decides, or intervenes at defined points rather than delegating the entire outcome to a model.
A change in model behavior or performance as data or conditions differ from those used during development and validation.
A payment-stage decision about whether a transaction may proceed. Approval does not necessarily mean final settlement.
A system-to-system notification sent when an event occurs. Operations must account for delays, duplicates, and delivery failure.
A design property that lets the same request be retried without creating an unintended duplicate financial outcome.
A recording model in which every financial event produces balanced debit and credit entries across accounts.
A protocol-specific indicator of collateral safety. Crossing a defined threshold may make a borrowing position liquidatable.
A process that reduces or closes an undercollateralized position according to protocol rules, often with an incentive for the executor.
The difference between an expected trade price and the executed result, often influenced by size, liquidity, and market movement.
The point at which a transaction is considered practically or legally irreversible. Meaning and timing vary by system.
Among cases a model flags as positive, the share that are actually positive. Useful when false alarms create material cost.
In model evaluation, among all actual positive cases, the share the model finds. Useful when missed cases are costly.
A designed outcome in which a model does not decide because evidence or confidence is insufficient, routing work to another path.
Independent assessment of a model's conceptual soundness, performance, implementation, limitations, and controls.
A control pattern in which one person initiates or prepares an action and another independently reviews or approves it.
A defined expectation for service performance or response. An SLA should have a measurable clock, scope, and owner.
The underlying condition that produced an issue. A repair clears the current item; root-cause remediation reduces recurrence.
The risk that remains after controls or mitigation. A strong handoff names it rather than implying every issue is eliminated.