projects
production-grade ai systems. real problems, real data, real scale.
High-throughput URL shortener with Redis caching and 30-day auto-expiry.
URL shortener with a Redis hot-path cache (30-day TTL, refreshed on every hit) and SQLite persistence. Redirect flow checks Redis first — cache miss falls back to DB and repopulates the cache. Stats updates on redirect run in a background thread to keep latency near zero. Inactive links auto-expire via an hourly daemon thread; also runnable as a management command. Built with Django, redis-py, and vanilla JS.
Ask your database anything in plain English — get SQL, charts, and narrative reports.
Business intelligence platform powered by a 5-node LangGraph agent (planner → SQL gen → executor → analyst → narrator). Supports CSV/Excel upload and direct DB connections. Auto-selects chart type, detects anomalies, and generates plain English narrative. Conversational follow-up with persistent memory. Built with FastAPI, PostgreSQL, Celery, React, and RAGAS eval pipeline.
Feed your sales data — get demand forecasts, reorder alerts, and supplier drafts.
Inventory intelligence engine using Prophet for demand forecasting and XGBoost for anomaly detection. LangGraph orchestrates the full reorder workflow — detect low stock → forecast demand → calculate quantity → draft supplier email → await approval. Supports CSV upload and Google Sheets. NLQ over inventory data. Built with FastAPI, Celery Beat, PostgreSQL, and React.
Describe a market or company — get a real-time intelligence report from across the internet.
Autonomous market research agent that searches the internet for real events, news, filings, social signals, and hiring trends — then synthesizes a structured intelligence report with source citations. Covers executive summary, event timeline, sentiment trend, competitor signals, and risk/opportunity breakdown. Supports on-demand reports and continuous monitoring with alerts. Built with LangGraph, web search tools, RAG, FastAPI, and React.