I'm a current UVA Darden student. I'm building interactive tools I use to drill dense technical material — spaced repetition, generative problem sets, and live playgrounds instead of static notes or slide decks. I'm also building tools with real-world use case based on my academic and professional experiences in ai x bio/longevity, ai x blockchain, VC and PE, etc.
The AI Stack — an interactive map of the AI industry from silicon to application layer, with ~20 company profiles, three money-flow traces, and a personal AI-cost calculator (subscription vs. pay-as-you-go, at your own usage). Live
TraceHound — an agentic on-chain crypto hack tracer: live hop-by-hop tracing from compromised wallet via the Etherscan API, cross-references a watchlist populated with real, current OFAC-sanctioned addresses pulled from Treasury's SDN list, LLM narration, and drafts a demand letter. Live
Deal Docket — a deal-sourcing dashboard built around an AI-enabled service-roll-up thesis, with a five-box scoring framework whose weights you can drag live to re-rank a 30-deal pipeline in real time. Live
PhaseSignal — pulls real, live trial data from the ClinicalTrials.gov API and scores each one against a cited published base rate, adjusted by four computed factors with a full live-reweighting breakdown — no black-box score, every number sourced or fetched. Live
IB Technicals Fluency Trainer — a merger-model cockpit, purchase-price allocator, and DCF sensitivity heatmap built as live playgrounds, backed by generative math drills and spaced repetition.
Consulting Case Prep Trainer — a profit-diagnosis game that scores hypothesis-driven reasoning under a limited information budget, not memorized frameworks, plus a market-sizing builder and exhibit reader.
The Operator's P&L Room — an eight-quarter run-the-business simulator under real leverage and covenants, paired with a 13-week cash-crisis room for distress-operator decision-making.
Most are self-contained apps — vanilla HTML/CSS/JS, no framework, no build step, no dependencies — designed, built, and shipped solo end-to-end: data model, interaction design, and deployment. The AI Stack, Deal Docket, and PhaseSignal split data from rendering (data.json + app.js) so their content can update without touching the app itself; the three trainers are single-file by design since their content doesn't go stale the same way. PhaseSignal goes a step further with a real Python data pipeline (data/build_dataset.py) that pulls and scores live data rather than reading from hand-authored fixtures. TraceHound is the exception — it needs a real backend to keep API keys server-side, so it's a Next.js app with server-rendered API routes, deployed on Vercel.