Skip to main content
Portfolio Picture

About Me

I am Nithin Varghese, an AI/ML consultant, product engineer, and applied AI engineer building practical systems with AI agents, multi-agent workflows, LLMs, fine-tuning, LangGraph, deep learning, and production computer vision. My work sits at the intersection of production AI, real-time video analytics, edge deployment, and human-centered product design.

I have spent the last several years shipping computer vision systems into demanding real-world environments, and I now focus on building AI products that turn voice, text, images, and workflows into useful software experiences.

Professional Experience​

AI/ML Consultant & Product Engineer (Dec 2024 - Current)​

Independent, New Delhi, India

  • Built latentstory.com, a filesystem-centric LangGraph multi-agent pipeline that generates illustrated storybooks and print-ready PDFs with realtime streaming UI and persistent workspaces.
  • Built and shipped FlowWriter, a voice-first Socratic ideation tool that converts conversations into structured outputs such as blogs, tweets, SCAMPER explorations, and product notes using LLM agents.
  • Work across LLM agents, RAG, product prototyping, workflow design, and AI-assisted software development.

Head Computer Vision Engineer at DeepSight AI Labs (Feb 2021 - May 2024)​

DeepSight AI Labs, Gurgaon

  • Led end-to-end delivery of multi-site computer vision systems across mining, energy, banking, and retail, owning scoping, architecture, model operations, and on-site rollout.
  • Built production video analytics pipelines for Coal India sites, covering no-PPE detection, fire detection, intrusion, crowd and vehicle counting, tailgating, idle time, and zone-breach detection.
  • Delivered safety and operations systems including ArcelorMittal person-under-steel-roll detection, national exam proctoring analytics scaling to 10,000+ cameras per exam, ATM security analytics, and HPCL wagon tracking/counting.
  • Architected edge-first realtime analytics for banking and retail deployments, optimizing throughput, latency, and reliability across field hardware.
  • Deployed facial recognition systems with vector database retrieval from training through indexing and site rollout.
  • Spearheaded hardware-aware inference optimization using TFLite, TensorRT, and OpenVINO for Raspberry Pi, CPU, and GPU deployments.

Computer Vision Engineer at DeepSight AI Labs (Sep 2018 - Feb 2021)​

  • Developed a camera tamper detection algorithm for CCTV monitoring systems.
  • Trained and deployed models for masked-face detection, helmet detection, and crowd analytics.
  • Supported production computer vision use cases from model training through deployment.

Freelance Machine Learning Engineer (May 2018 - Aug 2018)​

  • Developed a receipt information extraction system using YOLOv3-based object detection and OCR to transform receipt images into structured CSV outputs.

Certifications & Professional Development​

  • Mastering LLMs for Developers & Data Scientists: upskilled in RAG, LoRA, fine-tuning patterns, and practical LLM application workflows.
  • Solve It With Code (fast.ai): mentored by Jeremy Howard; strengthened AI-assisted coding and dialog-engineering workflows.
  • Hugging Face Agents Course: certified after securing a leaderboard position for an LLM-based agent on the GAIA benchmark.

Skills​

  • Computer vision and video analytics
  • Deep learning and machine learning systems
  • PyTorch, TensorFlow, TFLite, TensorRT, OpenVINO
  • LLM agents, RAG, LoRA, fine-tuning workflows
  • LangGraph, agentic workflows, AI-assisted coding
  • MLOps, model deployment, edge inference, system design
  • Python, Bash, Linux, Windows
  • Cloud platforms including AWS, GCP, and Azure

Get in Touch​

Feel free to reach out to me to discuss opportunities across AI agents, LLM systems, fine-tuning, LangGraph, applied AI, and computer vision: