Asmit Verma

I’m an AI/ML Engineer working on production-grade GenAI systems—especially RAG pipelines, multi-agent workflows, and end-to-end MLOps. I care about shipping solutions that are reliable, measurable, and deployment-ready.

At Coforge as a Trainee Engineer—AI/ML, I engineered a Financial Market Intelligence RAG pipeline using CrewAI, LangChain, FastAPI, and OpenAI/Hugging Face models with FAISS vector search and semantic retrieval. I also designed autonomous multi-agent workflows (Retriever, Analysis, Portfolio, Risk Assessment) to automate financial analysis, reasoning, and task orchestration. To improve performance, I implemented hybrid search, embedding optimization, prompt engineering, retrieval validation, conversational memory, and feedback-driven response refinement.

In my projects, I architected and shipped a Production-Grade Multimodal RAG System: a 13-module pipeline with hybrid retrieval, reranking, citation verification, and automated evaluation over real PDF reports. I built an extraction layer for complex tables using PyMuPDF and pdfplumber, and I added an evaluation harness for retrieval hit rate, citation faithfulness, refusal correctness, and latency to catch regressions early.

I’m especially motivated by evaluation-driven GenAI engineering—turning prototypes into robust deployments with clean APIs, CI/CD, and monitoring. I enjoy combining model orchestration, grounded generation, and practical MLOps workflows so teams can trust the outputs.