Real projects where I solved actual problems and turned data into decisions that drive business value
Built independently outside of coursework
A B2B sales tool for data/tech consultancies targeting small trucking carriers. Instead of spending weeks building a custom demo before a client commits, you drop in their data and the pipeline autonomously produces three deliverables: a Plotly insights report, a public-facing company profile, and a live Streamlit ops platform — all deployed over HTTPS with no manual steps.
A 7-agent autonomous build loop that grew a SaaS prototype over 146 cycles — each cycle plans a feature, writes the code, runs QA, and self-deploys. The result: a working trucking business management app with auth, Stripe subscriptions, AI receipt scanning, and 27 shipped features — built without a line of manual code.
A self-hostable job application tracker where every interaction—adding listings, updating statuses, recording notes—happens entirely through conversation with an embedded Claude AI assistant. Real-time dashboard with a live-updating table, rich detail panel, and color-coded alert callouts. Built on FastAPI with a vanilla JS frontend.
Completed during TripleTen AI/ML Engineering bootcamp
Analyzed 35+ years of gaming industry sales data to forecast 2017 product success across platforms, regions, and genres. Built statistical models to identify success patterns, then translated complex findings into clear marketing recommendations that non-technical teams can act on immediately.
Applied statistical hypothesis testing to analyze telecom customer behavior across 5 datasets. Converted technical statistical outputs (p-values, confidence intervals) into a clear executive recommendation on which calling plan to prioritize—bridging technical analysis and business decision-making.
Analyzed 30 million grocery transactions by systematically connecting 5 datasets to uncover customer shopping patterns—purchase timing, product pairings, and reorder behavior. Delivered insights in a Jupyter notebook designed for product managers to understand and act on immediately.
Analyzed streaming platform data to guide content acquisition decisions. Systematically cleaned messy data (missing values, inconsistent formats, duplicates) and built reusable analytical functions—demonstrating engineering practices that scale beyond one-off analyses to production-ready solutions.