Samir Mishra

About Samir Mishra

I am a research-driven AI Engineer and Machine Learning Researcher who bridges the gap between rigorous mathematical theory and production-grade system design. My expertise spans the full AI lifecycle—from deriving first-principles algorithms (Bayesian modelling, GNNs) to architecting autonomous Agentic AI systems and optimising distributed workloads on custom hardware.CURRENT FOCUS: AGENTIC AI & SYSTEM DESIGN Operating as an Independent Researcher, I am currently pushing the boundaries of Agentic AI. I move beyond simple LLM wrappers to architect autonomous multi-agent ecosystems using LangChain, LangGraph, and CrewAI.Architectural Patterns: I implement advanced patterns like Orchestrator-Worker and Routing to solve complex, multi-step problems.Recent Work: Built a \"Market Research Orchestrator\" leveraging the Model Context Protocol (MCP) and Asyncio, applying the RESHADED system design framework to ensure fault tolerance and scalability.PROFESSIONAL EXPERIENCE At Intel Corporation (Habana Labs): I focused on the intersection of software and hardware, optimising Deep Learning workloads on custom Gaudi accelerators. I drove a 13% reduction in inference time for Generative AI models through mathematical optimisation of RoPE embeddings, KV-caching, and low-level kernel tuning. My work involved distributed model execution, runtime profiling (Perfetto/Grafana), and building robust CI/CD pipelines.At Siemens Technology: I played a key role in developing SDeRM, an award-winning decision recommendation engine. I fused Graph Neural Networks (GNNs) with Knowledge Graphs (Neo4j) to achieve 82%+ accuracy in industrial recommendations. This research resulted in a peer-reviewed publication on neural subgraph matching, demonstrating my ability to combine structured graph topology with unstructured semantic reasoning.CORE STRENGTHSMathematical Foundations: I have deep experience in Bayesian modelling, hypothesis testing, and probabilistic reasoning. I apply principled statistical inference and generative priors to handle uncertainty in both structured and unstructured datasets.Advanced AI: Expertise in Generative AI (LLMs, VLMs, Diffusion Models), NLP, Reinforcement Learning, and RAG/Graph-RAG pipelines.System Design & MLOps: I focus on production readiness. Whether using Docker, Kubernetes, or GitHub Actions, I design systems that are scalable and user-centric. I am proficient in deploying interactive front-ends (Streamlit) backed by robust APIs (FastAPI).Tech Stack: Python, C++, PyTorch, TensorFlow, NetworkX, LangGraph, Neo4j, AWS, SQL/NoSQL.

Experience

1. Ai Test Engineer

   IBM

   Mar 2026 — Present · IN

Education

• International Institute of Information Technology Bangalore

  Master of Technology - MTech, Computer Science

• Savitribai Phule Pune University

  Bachelor of Technology - BTech, Computer Engineering