Gaurav Sharma

I’m a Backend / AI Engineer with 2+ years shipping production AI and data systems end-to-end, and I’ve often operated as the sole engineer driving critical features. At LayerFive, I built a LangChain agentic system for conversational marketing intelligence used by 5+ clients, including an 8-skill modular architecture, dynamic mid-turn skill activation, and a registry of 39+ tools across ads, sales, visitors, customers, memberships, dashboards, and reporting.

I focus heavily on reliability and usefulness in production: I shipped real-time SSE streaming with Pydantic-validated structured output (text/chart/table/action blocks) and heartbeat keep-alives for responsive chat experiences. I also delivered self-serve reporting and dashboards for 12+ clients—combining a configurable report builder (comparisons, variance, formula metrics, rule-based dimensions) with a drag-and-drop dashboard backend (multiple widget types and per-user sharing controls). On the data side, I extended marketing-data attribution by shipping server-side Meta Conversions API (CAPI) tracking via the Stape Signals Gateway for durable, cookieless conversion attribution.

Before that, at InfoAnalytica, I integrated OpenAI GPT-4o into a Python pipeline for file-driven row-as-parameter inference (supporting synchronous and batch modes), and I implemented a Celery + Redis asynchronous system on a Django backend to orchestrate 44 scraping modules. I also built supporting products like a Next.js monitoring dashboard and a WooCommerce data extractor using asyncio/aiohttp, with significant EC2 cost and memory reductions. In parallel, I explore technically ambitious problems—like ImmunoRAG, a progressive 8-level RAG system engineered to run fully locally on CPU-only hardware—because I like building systems that are both effective and operationally practical.