Selected Work
Markets, research infrastructure, prototypes, and competition systems, grouped around the artifact, the outcome, and the public link when it can be shared.
Highlights
Short, visual snapshots for the outcomes that deserve to be seen first: live trading, placement context, and the artifact behind each result.
HUQT / HUTC Win
Traded eight live derivative contracts around a MarioKart Grand Prix, updating fair values across all four races while new rules and state changes hit the market in real time.
IMC Prosperity
Finished #193 overall with 487,287 XIRECs, #54 in America, and #56 in the final algorithmic round after round-by-round strategy, risk, and review work.
Portfolio
Shipped prototypes, research tooling, and market systems, with public links for deeper inspection when source or live deployments can be shared.
A public Python CLI that normalizes Kalshi market pulls, local files, model checks, sizing experiments, and post-run diagnostics into one repeatable workflow.
A lyric-recall game where the singer cannot see the words, then Spotify playback, Whisper transcription, and comparison logic score what they remembered.
Round-by-round strategy work built around simple market-making rules, product signals, voucher valuation, position controls, and review notebooks.
A Python experiment workflow for simulating noisy multi-objective results and showing how much trust to put in a Pareto frontier.
A time-ordered sentiment and market-data pipeline for comparing next-day equity direction classifiers without leaking future information.
Shared Python wrappers for loading dialogue turns, entities, ratings, and metadata across conversational recommendation datasets.
A small focus tracker centered on three repeated actions: track a session, review recent work, and adjust habits without cluttering the main flow.
A browser SQL practice app with a local query playground and a Flask server check that rewards real query logic over client-side shortcuts.
Several projects point directly to source repos or live deployments so reviewers can inspect the work instead of only reading summaries.
Repos + LivePareto bands and SpokenCRS are framed around reusable tooling: experiment setup, common schemas, and outputs that make model comparisons easier to trust.
Reusable ToolsThe common thread is practical tooling: make raw data, model output, or practice feedback easier to inspect and act on.
Tools First