I am Himadri Mishra, a senior AI engineer building agentic workflows, LLM systems, ML infrastructure, search, and computer vision products.
LLMs are one component. The hard part is execution, evaluation, recovery, cost, and artifacts that teams can trust.
- Flagship AI systems case study: Agentic market research platform, DAG execution, sandboxed code, judge verification, charts, Deck IR, and PPTX automation.
- Platform depth: ML infrastructure rescue, cost reduction, pod reduction, faster builds, and reliability boundaries.
- ML product depth: Computer vision product systems, real-time CV under device, latency, and UX constraints.
- Interview surface: Interview me, source-grounded answers on architecture, tradeoffs, operating style, and risks.
himadri.dev, evidence-first portfolio for production AI systems work.handwrite-font-maker, converts handwriting specimen sheets into installable fonts.NTM-One-Shot-TF, older one-shot learning implementation that still attracts ML interest.Botnet-Detection-using-Machine-Learning, older applied ML project with long-tail usage.
- Explicit execution graphs beat free-form agent chaos.
- Artifact evals matter more than prompt evals alone.
- Cost, latency, retries, and observability are product architecture.
- Good AI products expose intermediate state without leaking private data.
- Computer vision and infrastructure work keep my LLM systems grounded in real production constraints.
Portfolio · Case studies · Resume · LinkedIn · Email