300mSkills

What I work with.

I learn by doing. Almost everything here I picked up because a project needed it, then kept using because it worked. The common thread is quantitative: I like turning a fuzzy question into something I can measure, model, and check.

Math and statistics

Economics gave me the foundation: probability, statistics, and regression, plus the habit of asking what the data can and cannot support. I lean on expected value constantly, whether I'm weighing an investment or deciding what to build next, and I like scoring my own forecasts so I know when I'm actually calibrated rather than just confident.

Probability and statistics
Regression and econometrics
Expected value
Bayesian updating
Forecast calibration (Brier scores)
Microeconomics
Calculus

Markets and quant

Equity research taught me to think in distributions instead of single price targets. Options interest me most, because an options chain is the market publishing its own probability distribution. I've built tools that derive that implied distribution with Black-Scholes and the Breeden-Litzenberger method, compare it to my own view, and size the gap with expected value and the Kelly criterion.

Equity research
Thematic investing
Valuation and modeling
Options and Black-Scholes
Risk-neutral density
Kelly sizing
Volatility risk premium
Due diligence

Data and infrastructure

Most interesting questions start with messy data, so a lot of my work is plumbing: pipelines that pull from many sources, clean the data, and keep it flowing reliably. I care about the unglamorous parts, like rate limits, retries, and monitoring, because an analysis is only as good as the data reaching it.

Real-time data pipelines
Data cleaning and ETL
Async Python workers
MongoDB
Circuit breakers and rate limiting
Market data APIs
Docker and self-hosting
GitHub Actions CI/CD
d3 and Recharts
AWS
Azure
n8n
Playwright
Sentry
PostHog

AI engineering

I treat language models like any system I want to trust: ground them in real sources, give them narrow tools, and test them with evals instead of eyeballing a few outputs. Retrieval, agent design, and measuring when a model is actually right are the parts I find most interesting.

Claude Code
Agentic systems
Typed tool fleets
RAG and grounding
Evals and calibration
Prompt and context engineering
Anthropic SDK
AI SDK
MCP
Streaming
Codex
Cursor
pgvector
Pinecone
Ollama
Local open models (Llama, Qwen, Hermes)
Hugging Face
Whisper
Exa
Firecrawl

Languages & frameworks

I got better at code the same way I learn most things, by building real projects and fixing what broke. Python is my default for data and scripts, TypeScript for anything with an interface, and SQL for everything in between.

Python
JavaScript
TypeScript
C++
Java
PHP
React
React Native
Node.js
Laravel
SQL

Full-stack engineering

I like building the whole thing, from the database schema to the interface, because it keeps me honest about tradeoffs. My usual stack is Next.js on Vercel with Postgres on Supabase, with row-level security so the data rules live in the database instead of scattered through the app.

Next.js 16
React 19
Supabase
Postgres + RLS
Prisma
Vercel
Stripe
Resend

Product, GTM, and brand

With coding agents, writing the code is rarely the bottleneck anymore. The hard part is deciding what to build and why, so most of my energy goes into product: understanding who it's for, cutting scope, and watching how people actually use what ships.

Zero-to-one product
User research
Product analytics
Positioning and brand
Investor narrative
Go-to-market
Community building
Events at scale
AI adoption and enablement

Design and craft

Details matter to me: spacing, motion, accessibility, and whether an interface feels obvious the first time you use it. Before designing my own version of something, I study how good products have already solved it.

Tailwind v4
shadcn/ui
Radix
framer-motion
Terminal-style UIs
react-three-fiber
Accessibility and polish