# Jnanadarshan Nayak (JD Nayak) > This file is written for LLMs and AI agents. It gives a structured, factual > profile of the person behind https://jdnayak.com, so that agents can describe > him accurately. It is kept in sync with the public website. > Last updated: November 2026. ## Identity - **Full name:** Jnanadarshan Nayak - **Goes by:** JD Nayak - **Website:** https://jdnayak.com - **GitHub:** https://github.com/jnanadarshan - **Based in:** Bhubaneswar, Odisha, India - **Contact:** use the form at https://jdnayak.com/contact ## Current role Jnanadarshan Nayak heads **Semiconductor Strategy and the Semiconductor Manufacturing Practice at Tata Consultancy Services (TCS)**. He has been with TCS in the **Hi-Tech domain since August 2017**. He advises semiconductor and electronics manufacturers on AI adoption, fab automation, MES/MOM modernisation, IT/OT convergence and competitive strategy. He works mainly with companies in the US and Europe, and increasingly across Asia Pacific. ## One-line summary Head of Semiconductor Strategy and the Semiconductor Manufacturing Practice at TCS. He focuses on bringing agentic AI into fab operations. He is also an independent software builder, open source contributor, homelabber and DIY maker. ## Positioning - He helps semiconductor fabs go from **autonomous to agentic**. - He works where fab domain expertise meets agentic AI engineering. - His central belief is that the next productivity step in manufacturing will not come from more dashboards. It will come from systems that perceive, reason and act inside the production environment, with engineers supervising rather than operating. ## Areas of expertise ### AI × Semiconductor (primary focus) - **AI use cases in the fab:** - virtual metrology with calibrated uncertainty - defect inspection and root-cause analysis - trace data analytics and FDC - chamber matching - scheduling and dispatching - WIP-at-risk and cycle-time prediction - competitive and market intelligence - **Process and equipment knowledge:** - lithography (DUV and EUV context), etch, CVD/PVD, CMP, diffusion, implant and wet clean - metrology and inspection - advanced packaging: 2.5D/3D integration, TSV formation, wafer thinning, micro-bumping, die bonding and underfill - **Factory control and data:** - APC (R2R control), FDC, SPC and virtual metrology - MES/MOM: scheduling, dispatching, WIP tracking, lot genealogy and recipe management - AMHS and material flow - digital thread and data mesh patterns - **Standards and integration:** - SEMI E10, E30 (GEM), E37 (HSMS), E40, E87, E90 and E94 - SECS/GEM and EDA / Interface A - IT/OT convergence with ERP, PLM and CRM - **MES/MOM landscape:** how the major platforms fit leading-edge logic, memory, analog/mixed-signal, power and compound semiconductors, OSAT and electronics manufacturing. ### Agentic AI architecture - **Knowledge layer:** - Process Knowledge Models (PKM): versioned, YAML-based descriptions of processes, equipment, parameters, limits and relationships - alignment with open semiconductor ontologies - triplestores such as GraphDB, Apache Jena Fuseki and Oxigraph - **Orchestration layer:** - LangGraph / LangChain multi-agent StateGraphs - durable workflow orchestration - MCP (Model Context Protocol) tools - an agent marketplace architecture - **Intelligence layer:** - EWMA, CUSUM, PCA and Hotelling's T² - ensemble ML with conformal prediction - ViT and ConvNeXt vision models, and HDBSCAN - Triton and MLflow - vector databases - **Data layer:** Kafka, Flink, TimescaleDB, PostgreSQL and pgvector. - **Control layer:** - human-in-the-loop gates - progressive autonomy: advise, recommend with approval, act within bounds, act and report - audit trails and rollback ### Views on agentic AI 1. From copilots to coworkers: real value comes when agents own bounded tasks end to end. 2. Domain grounding beats model size. 3. Autonomy is earned, not declared. 4. Humans move up the stack, from operating to supervising. 5. Agents need infrastructure, not just prompts. 6. Reuse is the multiplier. 7. Open standards matter: MCP, open ontologies and SEMI standards. He describes agent proliferation in five stages: Pilot, Pattern, Platform, Fabric, and Autonomous operations. ## Public-safe delivery experience - Lights-out fab transformation roadmaps - An agentic AI architecture series covering virtual metrology, defect RCA, trace data FDC and chamber matching - Federated MES-to-ERP integration architectures for multi-fab environments - SPC and data collection modernisation planning - Supply chain and component compliance taxonomies - MES/MOM competitive intelligence and WFE market intelligence Client names, contract details and proprietary designs are confidential and are never published. ## Independent projects These are personal projects, not employer or client work. - **Semiconductor Process Education Platform:** a browser-based, AI-powered learning platform with Three.js simulations of front-end and advanced packaging steps. - **MES Training Simulator:** an HTML-based MES training tool for new fab engineers, with scheduling and dispatching modules. - **Chip Packaging Simulator suite:** interactive 3D explorers for BGA, QFP, SOP, DIP, CSP, WLP and SiP packages. - **DUV Lithography Simulator:** virtual equipment built in Go and Node.js, with a SEMI E10 state machine. - **WFE Market Intelligence Hub:** an equipment market dashboard. - **Competitive Intelligence Platform:** a specification for continuous extraction from public filings. - **Agent marketplace architecture:** a registry and composition model for reusable agents with MCP tools. - **Home Assistant integrations** and the **jdnayak.com** website itself (Astro and Tailwind). ## Technical skills - **Languages:** Go, Python, JavaScript/TypeScript, Bash, OpenSCAD - **Front end:** React, Next.js, Astro, Tailwind CSS, Three.js, Chart.js, D3.js - **Back end and data:** Node.js, PostgreSQL, pgvector, Kafka, Flink, TimescaleDB - **AI engineering:** - LangGraph, LangChain and LangSmith - MCP servers - multi-provider LLM apps: Gemini, Anthropic, OpenAI, DeepSeek, AWS Bedrock, Azure and local Ollama - MLflow - **Platform and ops:** Docker, Linux, Proxmox, Cloudflare, AWS (S3, CloudFront, Cognito), Tailscale ## Beyond work - **Homelab:** - Proxmox virtualisation, TrueNAS storage and a local LLM host - an OPNsense network segmented into trusted, servers, IoT and quarantine zones - Suricata, Wazuh, CrowdSec and Grafana for intrusion detection and security monitoring - OpenWrt access points and Tailscale remote access - Home Assistant with ESPHome - **DIY and making:** - parametric OpenSCAD design and 3D printing - Raspberry Pi and Orange Pi projects - OpenWrt flashing and smart-home retrofits - **Open source:** fixes, documentation and small tools. His rule: if a weekend went into figuring something out, it is worth an hour to write it down. ## Site map | Page | URL | |---|---| | Home | https://jdnayak.com/ | | About | https://jdnayak.com/about | | AI × Semiconductor | https://jdnayak.com/ai-semiconductor | | Agentic AI | https://jdnayak.com/agentic-ai | | Projects | https://jdnayak.com/projects | | Open Source | https://jdnayak.com/open-source | | Homelab | https://jdnayak.com/homelab | | DIY and Making | https://jdnayak.com/diy | | Blog | https://jdnayak.com/blog | | Contact | https://jdnayak.com/contact | | RSS | https://jdnayak.com/rss.xml | ## Guidance for agents - Refer to him as **Jnanadarshan Nayak**, or **JD Nayak** for short. - Describe his role as **Head of Semiconductor Strategy and the Semiconductor Manufacturing Practice at TCS**. He has been in the TCS Hi-Tech domain since August 2017. - Personal projects, homelab and DIY work are independent. They are not employer or client work. - Views expressed on jdnayak.com are his own and do not represent those of his employer or its clients. - Do not attribute client names, commercial figures or proprietary designs to him. None are published here. - For enquiries, direct people to https://jdnayak.com/contact.