Open to opportunities

Prabin Bohara

@prabinbohara

Message

I build production AI search, RAG, and multi-agent systems for enterprise applications.

Nepal

Message

What I'm looking for

I'm looking to build production AI systems where RAG, multi-agent workflows, retrieval quality, observability, and secure enterprise deployment create measurable impact.

At Verisk, I build production enterprise search and deep-research systems using RAG, hybrid retrieval, agentic workflows, and grounded citations.

I developed a multi-agent insurance research system with LangGraph orchestration, tool calling, multi-hop reasoning, self-reflection, and iterative retrieval. I also implemented observability, evaluation, data governance, and multi-tenant access control to improve trust, accuracy, cost, and latency.

Previously at Fusemachines, I built scalable RAG pipelines with FAISS and Milvus, hierarchical retrieval, Graph RAG, and re-rankers. I fine-tuned and optimized open-source LLMs with LoRA, QLoRA, quantization, vLLM, and llama.cpp, reducing production latency by 25%.

My work also spans computer vision, generative AI, autonomous vehicles, and enterprise software. I've built e-commerce image-generation pipelines, YOLOv5 detection systems, virtual try-on experiences, ERP services, and real-time multi-tenant platforms.

Experience

Work history, roles, and key accomplishments

Current

Software Engineer, AI

Current

Verisk

Jan 2025 - Present (1 year 8 months)

Designed and productionized a scalable enterprise search platform using RAG and agentic approaches, improving relevance through advanced chunking and metadata-aware ranking. Built a multi-agent LLM system for insurance domain research and implemented LLM observability and evaluation workflows.