Adarsh Pandey

Software Engineer

Noida, Uttar Pradesh, India

About

I'm a Software Engineer with deep expertise in Machine Learning. What I'm truly passionate about is bridging that AI expertise with interactive frontend development. I use my skills in Three.js and real-time graphics to bring computer vision models to life, creating immersive and user-facing applications.

Experience

  • AI ML Engineer at EnQuest PetroSolutions
    Dec 2025 - Present · 8 mos

  • Software Engineer at Deepmindz Innovations
    Mar 2022 - Present · 4 yrs 5 mos

    Architected the flagship AI coaching platform, using MediaPipe for real-time posture validation against video-based checkpoints. Enabled rapid content scaling by engineering a decoupled workout system with a JSON schema to separate instructional logic from application code. Delivered ML models with over 95% accuracy for yoga pose classification by building a full-stack training and deployment pipeline. Solved critical data scarcity by generating a dataset of 114,000+ synthetic images with an automated Blender and Python script. Achieved natural-sounding voice synthesis by engineering a generative model in PyTorch, trained on a 2+ hour dataset that I personally curated and cleaned from over 5 hours of raw audio.

  • Computer Vision Engineer at MACH India Private Ltd
    Jun 2021 - Mar 2022 · 10 mos

    Increased operational visibility for Tata Steel by deploying a CV system that monitored 8 concurrent camera feeds in their industrial yard. Improved data-streaming performance by 40% by re-engineering the ingestion module from Python to C++ with low-level socket programming. Automated vehicle and crane tracking by implementing fine-tuned YOLO-based object detection models.

  • Software Engineer ( 3D & Interactive Web ) at Minzor
    Dec 2020 - Jun 2021 · 7 mos

    Delivered a functional prototype for a web-based, interactive 3D avatar, pioneering a novel approach to user engagement. Owned the end-to-end 3D asset pipeline, from Blender modeling to optimised real-time rendering in the browser with Three.js. Achieved real-time interaction by integrating a conversational AI with response times of under 500ms.

  • Artificial Intelligence Intern at CODE GLOBAL PRIVATE LIMITED
    Sep 2019 - Jan 2020 · 5 mos

    Authored 5+ core Python modules for robotics, implementing actuator control, sensor interpretation, and a voice command interface. Selected by the company to deliver an AI & Robotics curriculum to students (grades 5-9) at Cambridge School, utilising LEGO Mindstorms EV3.