arham khan

At Samsung R&D Institute Bangladesh, I’m developing TypeScript backend features supporting the SmartThings iOS app. I’ve also set up servers and system-level infrastructure for development and testing, and contributed to security-focused backend reviews. On CiteKit, I built a LangGraph multi-agent pipeline that discovers, deduplicates, and ranks research papers. I paired LLM relevance scoring with cross-query frequency signals and developed a FastAPI backend for its research and citation features. I built Praxis’s event-driven backend for an AI career platform, using FastAPI, ARQ, Redis, and WebSockets to support CV parsing, job scraping, and application tracking. Its semantic job matching combines embeddings with skill overlap, while its career coach requires user confirmation before write actions. For the FictiPay Customer Churn Prediction project, I developed a model for approximately 850K mobile wallet accounts that achieved 0.98449 AUC-ROC on the private leaderboard. I engineered behavioral and transaction features with Polars, then tuned a LightGBM model with Optuna.