Engine 01 · Liquidity & Execution · Analytics & Automation
Cross-Asset Liquidity & Positioning Intelligence Engine: signal to execution, automated.
A desk-level execution intelligence framework that ranks when market-state changes should alter trade timing, order routing, and clip sizing; signals escalate only when magnitude, speed, crowding, and liquidity all converge simultaneously.
Architecture: Four fragility dimensions govern every signal: MOVE percentile (Normal <50, Watch 50–80, Escalate >80), depth change, spread state, and flow-risk classification (balanced / crowded / forced). The engine's decision outputs span UST/GoC, USD/CAD, WTI, and rates-vol; cross-asset interaction scores weight each signal for transmission risk before any execution decision is made. Single-indicator signals are filtered out as explicit false positives.
Workflow value: A four-phase implementation roadmap moves the desk from manual morning pack to an auto-refreshing Python/SQL pipeline with intraday alert routing, dashboard escalation, and a backtest-feedback loop for threshold tuning. Historical fragility reference points (Mar-2020 UST stress, UK LDI shock, vol-control unwind) calibrate the escalation ladder. The automation edge is not forecasting precision; it is reducing reaction latency and making execution rules explicit before gapped liquidity materializes.
Type
FICC Execution Intelligence
Signal engine · Escalation ladder
4-phase implementation
Python · SQL pipeline
Engine 02 · Market Data · Quality & Monitoring
Market Data Health Monitor: a market-data feed, validated before the desk trusts it.
A data-quality monitor on a PySpark pipeline that validates a market-data feed, models it into a star-schema lakehouse, and publishes SLAs with column-level lineage, so regressions get caught before downstream use.
What it does: the pipeline ingests the feed, runs validation checks against the expected schema and value ranges, and models the clean output into a star-schema lakehouse. Each published table carries an SLA and column-level lineage, so a consumer can see where every field came from and whether the feed met its freshness and quality bar.
Why a desk cares: trade reporting, reconciliation, and position monitoring all inherit the quality of the data underneath them. The monitor applies the same discipline a desk applies to orders and positions, continuous checks with clear escalation when something breaks upstream.
Type
PySpark pipeline
Star-schema lakehouse
SLAs · column-level lineage
Validation · regression checks
Study 03 · US Equities · Factor Research
Has the Value Premium Decayed? US equities, tested from the source.
A factor study of US equities that tests whether the value premium has weakened, built from the Ken French Data Library and published as a live dashboard that recomputes from the source file on refresh.
Method: the study works from the monthly HML factor series in the Ken French Data Library and estimates the premium across rolling windows and subperiods. Inference uses Newey-West (HAC) standard errors, so the conclusions survive the autocorrelation in monthly factor returns instead of overstating significance.
Why a desk cares: factor premia sit underneath equity positioning, and the evidence quality matters as much as the estimate. The dashboard pulls the same public source the study used and recomputes the analysis, so the numbers on screen stay current rather than freezing at publication.
Type
Ken French Data Library
Newey-West (HAC) inference
Live dashboard · recomputes from source
Python · ECharts
Engine 04 · Fixed Income · Rates & Portfolio Analytics
Tenor: Canadian rates, risk, and portfolio construction in code.
A fixed-income analytics engine, built from first principles in pure Python, that constructs the Government of Canada curve, prices a provincial portfolio, and produces the duration, key-rate risk, relative value, scenario P&L, and ESG allocation a rates desk runs on.
What it computes: a GoC par curve, then bond-level risk straight from the cash flows, modified and Macaulay duration, convexity, DV01, and key-rate durations at the 2y / 5y / 10y / 30y nodes, so the desk sees where on the curve the interest-rate risk actually sits rather than a single parallel number. Provincial bonds (Ontario, Quebec, BC, Alberta) are priced at their spread to Canada, with a relative-value screen flagging each issuer rich, fair, or cheap against the provincial sector.
Portfolio and positioning: it aggregates a roughly $212mm book to a weighted yield, modified duration, convexity, and total DV01 (about $181k per basis point), then fully reprices the book under parallel, bear-flattener, and bull-steepener curve moves rather than a linear estimate. A duration rebalance to target and a green-bond overlay complete it; the positioning note recommends trimming long-end duration, adding the cheap provincial on spread, and rotating long-end holdings into green 10s to cut duration and lift the ESG sleeve at once. North-star: a desk-ready positioning call an analyst can defend.
Type
Canadian rates · GoC and provincials
Duration · DV01 · key-rate
Scenario P&L · relative value · ESG
Pure Python
Note 05 · Cross-Asset Strategy · Macro Positioning
Cross-Asset Regime Transition: rates, oil, CAD, and liquidity positioning.
A probability-weighted institutional strategy note mapping regime transmission across US rates, USD/CAD, WTI crude, and volatility, built for desk-ready implementation at CIBC Global Markets FICC.
Regime read: Late-cycle inflation shock with policy divergence and fragile liquidity. Markets underprice both rates-vol persistence and liquidity fragility. This is not a clean CAD/oil bullish setup. The same oil move can be simultaneously CAD-positive (terms-of-trade), CAD-negative (USD/rates-vol channel), and duration-negative (inflation credibility) depending on which transmission channel dominates.
Core thesis: Trade the transmission path, not the headline. Best expression is conditional: CAD crosses over outright USD/CAD shorts, conditional steepeners over blunt duration, and event gamma over carry-only exposure. Asymmetry favors CAD short-squeeze (positive convexity if oil and BoC repricing align), belly duration convexity, and oil tail hedges via call-spread tails. A probability-weighted five-scenario matrix anchors the positioning framework: Sticky Inflation (30%), Growth Slowdown (25%), Commodity Shock (20%), Liquidity Tightening (15%), Risk-On Reflation (10%).
Desk
Rates · FX · Commodities · Vol
5-scenario probability matrix
May 2026
Framework 06 · Cross-Asset Strategy · Market Structure
Markets Thinking Framework: reflexive transmission, regime evolution, execution-aware decision-making.
The foundational mental model: how recursive feedback loops connect price, positioning, liquidity, and volatility, and what separates institutional-quality thinking from naive directional conviction.
Core standard: A view is not institutional until it specifies confidence decay, failure modes, execution path, and expression quality. Modern markets are dominated by a recursive loop: price moves change positioning, positioning shifts change liquidity depth, liquidity changes bind volatility limits, which alter execution capacity and ultimately drive price discovery. The loop is self-reinforcing and must be anticipated, not reacted to.
Structure: Regime migration is mapped across four states (Calm, Transition, Fragile, Stress) with scoring across volatility, liquidity, crowding, and reflexivity dimensions; the Fragile quadrant is maximum on all four. An execution-aware decision tree sequences every trade thesis through signal validity, falsification conditions, expression choice (option / RV / delta), and execution mode (passive / aggressive). Three failure-mode overlays (policy reaction shifts, liquidity vacuum, reflexive reversal) each carry explicit adaptive responses. The goal is not forecasting a single path; it is identifying what markets underprice, how shocks propagate, and which structure survives adverse path dependency.
Type
Cross-Asset
Regime migration · Reflexivity
Execution-aware decision tree
Failure-mode overlays
Engine 07 · Cross-Asset · Event & Macro Pipeline
Corroborate: a macro event pipeline read against price.
A cross-asset event record ingesting rates, policy and filings on a 15-minute cadence, plotted behind price and market cap so a move and the news around it sit on the same vertical.
Architecture: three lanes, macro, markets and micro, over 63,505 stored events. The macro layer pulls the 10-year and 2-year, fed funds, CPI and unemployment from FRED and ingests Federal Reserve policy releases directly. The micro layer reads SEC EDGAR, where the 8-K item codes arrive inline at no extra request cost and turn a filing from “Apple filed an 8-K” into a stated reason.
Ticker tagging is case-sensitive on symbols and case-insensitive on names, which is not cosmetic: thirteen S&P constituents have tickers that are ordinary English words, so matching without regard to case tags Allstate on the phrase “all eyes are on the Fed”. Symbol matching is skipped entirely when text runs more than 60% uppercase. The coverage universe expands as companies recur in events beside covered names.
Type
FRED · EDGAR · Fed releases
63,505 events · 15-minute cadence