Notes on things that resist neat answers.
About Me- I'm Haochi Zhang. I work in financial model risk and build applied AI systems.
- I'm drawn to problems without clean solutions — reliable systems around imperfect models, and research, analysis, and information workflows that don't reduce to a formula.
- I care about the unglamorous distance between a prototype that runs once and a system people actually rely on.
- This site is where that thinking lives — written across finance, AI systems, and building, in public.
Thinking about
How AI changes financial research workflowsBuilding reliable systems around imperfect modelsWhat makes a prototype become a usable product
Now building
Work
Model risk — financial & statistical models across risk, stress testing, and business planning. Previously quantitative advisory.
Recent notes All notes →
Why Most Agent Demos Stop Before the Hard Part
The gap between doing a task once and operating the same workflow repeatedly.
Don't Write Fallbacks
A fallback postpones an error to where it's harder to diagnose; orthogonal modules remove the class.
The MCP Surface and the Exec Surface
Why every LLM-generated tool call should cross one audited surface — and what stays in code.
List the Excuses an Agent Will Use to Skip a Step
Models don't cut corners; they generate a reason to. Name the rationalizations up front.
Financial RAG Is Not Just Document Search
Filings, market data, calculations, and provenance each want a different system layer.
Model Risk as a Knowledge Representation Problem
What's lost when a model inventory is treated as a spreadsheet, not a graph.
A Standard for Agent-Built Projects
Separate project facts, agent behavior, rules, and execution — before implementation begins.
Evidence That Can Be Traced
Traceability isn't a source column; it's every number reaching back to its raw fact.