Shubham Sinha

About Shubham Sinha

Generative AI, Microsoft, and AWS-certified Data Scientist and Machine Learning Engineer with 7 years of hands-on experience. Proficient in predictive modeling, data processing, and ML algorithms leveraging Python and PySpark. Expertise includes constructing and deploying generative AI solutions with OpenAI, adeptly employing Agile methodologies, and seamlessly executing data science projects. Skilled in integrating solutions across big data platforms for optimized performance.Demonstrated success in developing tailored data science solutions for complex business challenges, coupled with a track record of leading teams, mentoring junior data scientists, and driving project success. Certified in both AWS and Azure cloud platforms, with a knack for deploying solutions efficientlyNotably, spearheaded the development of numerous POCs and conducted solution demos alongside Managing Directors and Partners across a diverse clientele spanning pharmaceuticals, finance, telecom, and airline industry.

Experience

1. Artificial Intelligence Engineer

   IBM

   Apr 2025 — Present · Bengaluru, IN

   Agentic AI Platform (A2A Multi‑Agent System)Built an enterprise Agent‑to‑Agent (A2A) Agentic AI platform enabling multiple LLM agents to collaborate using shared memory, task delegation, planning, tool‑calling, and reasoning. Powered autonomous workflow execution across IBM App Connect, reducing manual steps and enabling scalable multi‑agent orchestration. Tech: Python, LangGraph, Milvus, FastAPI, App Connect.Enterprise RAG System (High‑Accuracy Retrieval)Designed advanced RAG pipelines using watsonx.ai, GPT‑4o, LLaMA‑3, and Milvus to deliver accurate, grounded enterprise search. Implemented hybrid retrieval, chunking, metadata filtering, reranking, and automated evaluations. Achieved 85–92% accuracy and 30–45% fewer hallucinations for document‑heavy workflows. Tech: Milvus, GPT‑4o, LLaMA‑3.Milvus Vector Search OptimizationDeveloped optimized Milvus vector pipelines with improved ingestion, deduplication, clustering, and indexing. Delivered 18–25% nDCG@10 gains and ~35% token cost reduction, enabling faster, more precise semantic retrieval for LLM workflows. Tech: Milvus, Python, Embeddings.MCP Tool‑Calling AutomationIntegrated Model Context Protocol (MCP) with GitHub, App Connect, and webMethods to enable autonomous workflow triggers, repo interactions, and decision automation. Resulted in end‑to‑end workflow automation and reduced operational overhead. Tech: MCP, GitHub API.LLMOps & Production DeploymentProductionized AI services using FastAPI, Docker, Kubernetes (AKS/EKS) with automated evaluation pipelines via LangSmith, PromptFlow, and Arize. Ensured >95% guardrail compliance, strong observability, and stable production performance.Pytest‑Driven CI AutomationBuilt a pytest-based CI suite covering unit, integration, data validation, and RAG regression tests. Reduced regression defects by 40% and system failures by 20–30%, enabling safer, faster releases. Tech: Pytest, GitHub Actions.

Education

• Bharati Vidyapeeth

  Bachelor of Technology - BTech, Electrical and Electronics Engineering

  2013 — 2017

Skills

• Image Processing
• Javascript
• Java
• Opencv
• Data Analytics
• Machine Learning
• Html
• Public Speaking
• 8051 Microcontroller
• Computer Vision
• Cascading Style Sheets (Css)
• C++
• Python