Lead Product Manager · New York
For high-stakes workflows. At Quantile (LSEG), that means agents that cut expert work from hours to minutes, and market infrastructure used by global banks. Outside work, I ship independent products end-to-end: RideLens, Daylight, RailDrop, and Gridiron.
Agents + markets · Quantile systems
Production AI · human gate · multilateral compression
Hours → to minutes
Schematic · Quantile systems · Offsets inside a risk envelope
Phases1-3 · select to hold
Reading the schematic interfaceTwo production agents, one for compression-run setup and one for incident triage, cut expert work from hours to minutes; their typed actions wait at a human approval gate; and, separately, offsets settle multilaterally through a central settlement service, with 34% more notional reduction per run.
Cross-currency compression · central settlement
Before: banks can only net in pairs · network offsets sit unused
+34%reduction per run vs bilateral
Schematic · eligible notional ≠ cash saved · 18 banks · 12 pairs
Phases1-4 · select to hold
Reading the schematic interfaceEighteen banks netting only in pairs, then a multilateral run settled through a central settlement service, compressed with trillions of dollars of notional made eligible and 34% more reduction per run, each book inside its constraints.
RideLens · compare board
Illustrative trip · four providers, three sort rules
4 → to 1providers → to one board
Illustrative quote types, not live fares · the final fare is always confirmed in the provider app
Phases1-4 · select to hold
Reading the illustrative interfaceOne trip mapped once on live roads across Uber, Lyft, Empower and Curb, and the same quotes ranked by price, soonest pickup or value, with upfront prices, ranges and estimates kept distinct and expired quotes never on top.
What I work across
Production AI, market infrastructure, and 0→1 product work
01Selected work
Two production AI agents that cut expert work from hours to minutes, and market infrastructure that unlocked previously ineligible notional.
A production AI agent inside a live operations workflow, not a side chatbot. Bounded actions, operator visibility, and zero AI-initiated configuration errors in its first six months.
When something broke in production, ops still paged senior engineers by default. The agent gathers domain evidence first, sharpens the diagnosis, and escalates only when the evidence is incomplete.
Banks held large cross-currency books that could only net in pairs. A central settlement path made network-wide offsets executable, so notional stuck in pairs became eligible to compress.
Also shipped
02Measured change
Each row is one system from the case studies: how the work ran before it shipped, and what changed after. Select a row to read the case.
All workFigures from each case study
03Labs
Consumer tools, a live football command center, and a native macOS app, each designed, built, and shipped live. No demos. Four of the eight have case studies here; the whole bench is on Labs.
01 / Consumer
Every ride, one comparison.
Live roads. Real rate cards. Uber, Lyft, Empower, and Curb ranked before you open four apps and guess.
02 / Native
Your screen, through the day.
Warmth and brightness on a schedule you set: gradual, offline, and honest about what it is doing and why.
03 / Consumer
Know when your train gets cheaper.
Book the trip. Watch every bookable Amtrak option across your window. One email when a listed fare actually drops.
04 / Live data
Every game. Every drive. One view.
A live NFL and college football command center: a 3D field for every game, with the reported ball spot, win probability, and odds beside it.
04Approach
Consequential workflows need context, bounded tools, deterministic services, human control, and evidence that the system deserves more scope.
Full approach01
Map handoffs, exceptions, and the cost of being wrong before choosing a model. If the operating model would not change, it is not an agent problem.
In the workSetup Agent
02
Context, bounded tools, deterministic services, and human control are the product. Autonomy grows one rung at a time, on evidence.
In the workIncident AgentAgent Platform
03
In markets and in agents, a late failure is a product failure. Validation, evals, and operator visibility belong in the design, not in the cleanup.
05Writing
Three short theses on AI product economics, autonomy, and infrastructure.
All notes01 · AI product economics
Routing intelligence is a unit-economics decision.
Read the thesis
02 · Agent controls
Evidence, reversibility, and control determine where an agent should be allowed to act.
See the approach
03 · Fintech × AI
Data quality, APIs, entitlements, and workflow design decide whether the model actually ships.
See the work
06Contact
Get in touch
I ship agents that change how real operations run and market infrastructure used by global banks, and I build independent products end-to-end. If that is the bar for your team, email me.
j / k sections & systems · Enter on a system · y link · t top · e email · ⌘K jump
Jump anywhere