Turning investment data into decisions people can trust.

"I went into finance to work on the decisions that actually shape people's lives."

How they save, invest, and plan for the future they want. What keeps me there is that I'm genuinely obsessed with problems. It's the same itch that had me decoding neural signals to control a robot arm in grad school, and building emergency-brake electronics for a Hyperloop pod. Different domains, same compulsion to take a system apart and rebuild it cleaner. Today that shows up as a couple of years managing data analytics & insights for a portfolio platform (equity holdings, FactSet workflows, dashboards for PMs and analysts), and now as a Sr. Finance Associate certifying the methodologies Fidelity uses to profile clients and allocate their assets. I've also become something of an internal AI advocate on my team, pitching tools rather than waiting for someone else to build them.

I picked up the CFA Institute Investment Foundations Certificate to strengthen the finance side of that equation, and I'm still working through AI-in-finance coursework. Off the clock, it's prompting techniques for LLMs, hiking, and shipping small personal web apps. A few of them are below.

Independent builds spanning investment tooling, applied AI, and native apps. Same habit as the About section: take it apart, rebuild it cleaner.

ENTRY NO. 01

Python · Streamlit · Compliance

Asset Allocation Certification Simulator

Built after spending my day job certifying real profiling and allocation methodologies, this is that same instinct in code: a 14-check certification engine modeled on real IPS constraints and MiFID II / SEC suitability rules, with confidence-scored triage (auto-approve / advisor review / human intervention) and a what-if analyzer for stress-testing recommendations before they ship. 34/34 tests passing.

View on GitHub →

ENTRY NO. 02

Backtesting · Interactive Demo

Crossover Downs

Seven rule-based trading agents race head-to-head on 10 years of real SPY / QQQ / AAPL / TSLA / GLD / BTC price history, with a pari-mutuel chip-staking mechanic. The backtests are real: Buy & Hold beats moving-average crossovers in 5 of 6 tickers, and the game's own "Form Guide" tells you so before you bet.

Demo available on request

ENTRY NO. 03

SwiftUI · iOS · On-Device AI

Tally

A native iOS expense tracker: type an expense in plain English, hit enter, done. Regex-based parsing, per-account budget pacing with alarm thresholds, and an on-device LLM (Apple FoundationModels) that cleans up notes and infers a spending category. No backend. Nothing leaves the phone.

Native app · not yet on the App Store

ENTRY NO. 04

1.38 Sharpe · 94% Hit Rate

Quant Research · Numerai Signals

Numerai Signals: RSI-Contrarian

Tested six candidate stock-ranking formulas against 20+ years of Numerai's real target data before touching an upload button. Fading extreme RSI (mean-reversion) held up in-sample and out-of-sample (Sharpe 1.38 / 1.16); pure momentum was consistently negative. Caveats documented too, since a well-known factor like this is unlikely to be orthogonal alpha.

Research complete · not yet staked

Happy to talk investment data, applied AI, or anything I've built above.