First-year B.Tech CSE (Data Science) student who ships tested developer tools, backend services and data-science projects. Deterministic behaviour, real test suites, written docs — that's the bar for everything below.
Portfolio · Email · Kadapa / Chennai, India · Open to SWE internships
| Project | What it is |
|---|---|
| relay-flags | Deterministic feature-flag engine in TypeScript — typed rules, percentage rollouts, append-only audit log, HTTP API + CLI, Docker, CI, 3 test suites. |
| ml-fraud-detection | Real-time fraud scoring: XGBoost/LightGBM ensemble with SHAP explainability, FastAPI inference, Kafka streaming, Prometheus metrics. |
| algo-trading-backtester | Event-driven backtesting engine — strategies, indicators, parameter optimizer, WebSocket dashboard. 7 test suites. |
| omnitrix | Terminal dashboard and theme engine for observing AI coding-agent sessions. TypeScript. |
Also: thanvish-portfolio (zero-dependency vanilla site) · python-in-22-days & java-in-22-days (interactive browser-based learning platforms) · dynamic-island (Tauri + Svelte + Rust desktop widget)
Every metric below is written to a reports/metrics.json by the code itself — nothing in these READMEs is hand-typed.
| Project | What it is |
|---|---|
| digit-recognizer-streamlit | Draw a digit in the browser, get a live prediction — MNIST neural net at 97.1% test accuracy, with center-of-mass canvas normalization so hand-drawn strokes match what the model trained on. |
| nlp-sms-spam-classifier | TF-IDF spam filter comparing Naive Bayes, LogisticRegression and LinearSVC — 0.966 F1 on held-out spam, scored on F1 rather than accuracy because only 13% of messages are spam. |
| ml-titanic-survival-prediction | Three classifiers compared under 5-fold CV with all preprocessing inside the sklearn Pipeline, so no test-set statistics leak into training. Best holdout accuracy 85.5%. |
| eda-retail-sales-insights | EDA over 1.07M real UCI Online Retail II transactions — KPIs, RFM segmentation and 6 charts. Finding: 13% of customers generate 55% of revenue. |
- DSA — arrays, trees, graphs, dynamic programming; consistent problem practice
- Backend systems — API design, caching, message queues, containerised deployment
- Data science / ML — feature engineering, gradient boosting, model evaluation and explainability
- Email — thanvishreddyco@gmail.com
- Portfolio — https://thanvish-portfolio.vercel.app