Antara Bhavsar

I’m a software engineer focused on turning real-world, ambiguous problems into production-grade RAG pipelines and agentic LLM systems. At CNS, I design system workflows and iterate based on usage and feedback, including FastAPI-based distributed backend services using Claude and local LLMs.

I’ve built tool-augmented LLM agents with LangChain/LangGraph (10+ tools), cutting latency by 35% and achieving >99% reliable inference through caching and fault-tolerant design. On AWS, I’ve delivered cloud-native distributed data pipelines for LLM and embedding-powered automation, improving semantic search and incident resolution by 30%, and using Amazon Q and Bedrock workflows to reduce manual effort by 60% while improving MTTR by 40%. I also value SDLC best practices and reusable modules, and I’ve contributed to high-throughput tissue mapping pipelines (30% throughput gain) and secured 2nd place at Motorola India Hackathon 2023.