Hi, I'm Ansh
Full-stack developer and AI/LLM engineer based in India. I deploy and maintain everything myself — from architecture to production.
Full-stack developer and AI/LLM engineer based in India. I deploy and maintain everything myself — from architecture to production.
Currently completing B.Tech (graduating 2027) — open to internships, part-time contract work, and full-time roles starting immediately.
I work across the full stack — but I specialize in three areas where I can go deep.
Production-grade APIs and real-time communication systems that handle concurrent users at scale.
Responsive, performant interfaces with modern React patterns, animations, and SEO-first architecture.
Multi-provider AI pipelines with structured prompting, streaming responses, and cost-optimised model routing.
Production systems I've built, deployed, and maintain — not class projects or tutorials.
Enterprise-grade AI orchestration engine that routes requests across 45+ LLM providers with task-aware intelligence and bulletproof failover. 100% OpenAI API compatible. Zero vendor lock-in.
Self-hosted productivity platform (Pomodoro, tasks, scheduler, team collab) built as an installable PWA. Deployed on Ubuntu VPS with Gunicorn, Nginx, and automated Certbot SSL.
Full-stack lead generation platform: Selenium scraper with anti-detection extracts 500+ listings into MySQL with automatic deduplication, surfaced through a React.js analytics dashboard.
Open-source VS Code extension that detects Python tracebacks in real time and streams AI-generated fix suggestions. Supports GPT-4, Claude, and local Ollama models interchangeably via LiteLLM.
GPS-verified student attendance with 300+ concurrent WebSocket connections at sub-50ms sync. Automated bulk-absent processing cuts admin workload by 80%.
Technical deep-dives on backend engineering, AI integration, and lessons from production.
PHP is often seen as purely synchronous, but modern libraries like ReactPHP and AmPHP unlock powerful asynchronous capabilities. This post dives into how event loops enable non-blocking I/O for high-performance applications.
I ran into an issue with our real-time attendance system where unreliable client-side GPS data led to false attendance records. The fix involved implementing server-side velocity checks to flag impossible student movements.
Prompt injection and adversarial attacks pose significant threats to LLM applications. I'll break down the engineering strategies I consider crucial for building secure, production-ready AI systems.
Currently completing B.Tech (graduating 2027) — open to internships, part-time contract work, and full-time roles starting immediately.
Interested in backend architecture, AI integrations, or full-stack projects that go to production — not staging. Let's connect.