Regulatory Change Radar
/ In progress · RAGA system that ingests versioned regulatory and sanctions documents, detects clause-level changes between versions, and answers grounded questions with citations to the exact source — and returns “insufficient context” when retrieval confidence is low. Python, FastAPI, a vector DB, and an LLM, tuned to never bluff. Still actively building.
open on github →
Migrating a compliance platform to Next.js
/ SDE InternLed the PHP → Next.js migration of the company site, wiring GraphQL alongside the existing REST APIs for leaner data fetching — 40% faster page loads and 90+ Lighthouse. Then built the unglamorous glue around it: Node APIs with caching and retries, and a Java/Selenium/Playwright crawler pulling 10,000+ compliance records a day from 50+ live sources.
open on github →
CRO Opportunity Engine
/ Shipped · LLMScrapes live Shopify store data — stock levels, reviews, descriptions — and uses an LLM to identify and rank conversion-optimization opportunities by impact, confidence, and effort. TypeScript, Next.js, OpenRouter, and Cheerio doing the reading.
open on github →
PySandbox
/ Shipped · InfraA secure Python code-execution sandbox: Docker-based isolation, resource limits, timeout handling, and a comprehensive test suite for the nasty edge cases. FastAPI in front, containers doing the dangerous part.
open on github →
A production site that moves at 60fps
/ Frontend InternBuilt the production website on Next.js 14 App Router with SSR and dynamic OG image generation, then made it feel alive: 20+ animated sections with GSAP and Framer Motion, holding a steady 60fps across devices. Shipped a reusable Tailwind component library that took dev velocity up 35%.
Dashboard modules, tightened
/ Frontend InternBuilt dashboard modules in React, Next.js, and TypeScript with REST integration, error handling, and caching. Refactored 15+ components and cut API calls by 40% through state optimization and request caching — the quiet kind of win that only shows up in the network tab.
The first thing I felt proud of — a chatbot
/ ML InternMy first internship: a conversational chatbot in Python with TensorFlow, NLTK, and Flask, reaching 89% intent-classification accuracy. Designed the SQLite schema and wrote the queries behind persistence and analytics. The first time something I built actually talked back.
← this is the thing. still being written.