Atharva User

I’m a GenAI Engineer with production experience at S&P Global, focused on building enterprise-grade LLM systems for energy and financial workflows. I work end-to-end across the GenAI stack, from rapid prompting and LLM API integration to reliable deployment patterns.

At S&P Global, I designed structured, few-shot, and chain-of-thought prompting strategies for document extraction across 600+ petroleum contracts—boosting extraction accuracy by 35% and reducing hallucination rates. I also integrated Azure OpenAI APIs into centralized fiscal data pipelines, combining LLM outputs with SQL/Spark logic and downstream backend services.

I prioritize dependability: I implemented reliability, consistency, and QA validation layers for AI-generated outputs, iterating on prompts, context windows, and retrieval strategies to improve factual grounding and reduce error rates. I also built PySpark + SQL + Python pipelines on Databricks for automated contract extraction and integration, saving 6–7 hours per contract in manual processing time, reducing manual research time by 90%.

Beyond LLM production systems, I enjoy multi-agent and research-style experimentation. I’ve engineered multi-agent orchestration with LangChain and LangGraph (specialized agents for retrieval, reasoning, synthesis), and developed RAG pipelines and evaluation-oriented workflows that help teams move faster—like reducing paper preparation time by 25%.