Skip to content
View dudailia's full-sized avatar

Block or report dudailia

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dudailia/README.md

Ilia Duda

Mathematics and Business Administration at Northeastern University, class of 2028. I work on quantitative finance and the systems around it — research stacks, market tooling, and applied LLM infrastructure.

Boston, MA · duda.i@northeastern.edu

What I build

closebooks-app — AI-assisted month-end close for CPA firms. Next.js, Supabase with row-level security, Anthropic Claude for transaction categorization, Stripe billing, QuickBooks Online journal-entry push. Sole engineer. Deployed, billing in test mode, no automated test coverage yet.

cricstate — A pre-registered negative result. I asked how much predictive signal exists in ball-by-ball cricket beyond match state, froze the decision rule before reading the test split, and measured. Match state captures ~93% of the recoverable signal; player identity adds +0.31%, below my own materiality bar, so the hierarchical model I intended to build was declined on the evidence. 16,754 matches, 4.75M deliveries, match-level bootstrap, leakage canaries in CI. The result is the refusal.

nucarbon — An estimation dashboard for the carbon cost of campus AI use, built in a week for the Northeastern Sustainability Incubator. Models rather than measures, and says so on every page.

startup-investment-analysis — Segment ranking over ~41k cleaned startup funding records. Course capstone, notebook in Russian, English summary in the README.

Background

Sole engineer at Glacier Capital Systems, building a Python research stack for a live options trading firm — scanner, multi-factor scoring, volatility surface, regime detection, and a real-time Next.js alert dashboard. Before that, daily market briefings on OFZ bonds and MOEX flows at BCS Bank's investment banking division, plus DCF and comparables work across energy and metals.

Working with

Python (polars, scikit-learn, scipy) · TypeScript / Next.js / React · PostgreSQL and Supabase · Anthropic API · SQL · Vercel

Notes

I try to publish negative results and known limitations alongside the work. Every README here states what is unfinished, untested, or simulated. If a repo says something is in production, it means deployed and reachable — not that it has customers.

Pinned Loading

  1. closebooks-app closebooks-app Public

    AI-assisted month-end close for CPA firms. Next.js 14, Supabase RLS, Claude transaction categorization, Stripe billing, QuickBooks Online journal push.

    TypeScript

  2. nucarbon nucarbon Public

    Estimation dashboard for the carbon cost of campus AI use at Northeastern. Models AI energy draw from published coefficients and stated assumptions.

    TypeScript

  3. cricstate cricstate Public

    A calibrated, leakage-audited benchmark for predictive signal in ball-by-ball cricket beyond match state - a negative result.

    Python

  4. startup-investment-analysis startup-investment-analysis Public

    Segment ranking over ~41k cleaned startup funding records. Data-analysis course capstone; notebook in Russian, English summary in the README.

    Jupyter Notebook