Shashank Shekhar

AI/ML (CV, NLP, RL) Engineer

About

ML Engineer with more than 5 years of experience in Deep RL, Computer Vision, NLP, Generative AI, LLM, and Video Analytics. I have a strong interest in autonomous driving and AGI research. I'm driven by curiosity and passion for continuous learning. Most of the time I like to work every waking hours that's why I thrive in the fast-paced, dynamic environment of startups. Open to Startup Opportunities.

Experience

Startup: Building and Iterating on AI ProjectsJAN 2023 - Present

  • Trained and fine-tuned various language models (LLMs), diffusion models (T2I) and audio models.
  • Successfully implemented a range of foundation models from scratch, spanning audio, image generation, and language understanding tasks.
  • Adapt pre-trained LLMs to specific domains and tasks using targeted datasets. Experiment with fine-tuning techniques to enhance LLM performance.
  • Design and implement Retrieval-Augmented Generation (RAG) systems, integrating LLMs with company knowledge base.
  • Optimize LLM architectures, hyperparameters, and training strategies to improve accuracy, efficiency, and reduce bias.
  • Solid understanding of data processing pipelines and techniques for working with large images and text datasets.
  • Extensive experience using PyTorch to train and deploy machine learning models. Deep understanding of various ML Framework core concepts and features.

Honeywell India (Impact brand), Pune - Senior ML EngineerJuly 2022 - Nov 2022

  • Deployed light weight Object Detection model on Amlogic Neural Hardware Accelerator.
  • Implemented a novel camera tampering algorithm using the OpenCV C-API and integrated it into the existing codebase.
  • Helped simplify and debug the entire 20k line of C code.

Startup Idea, Distributed Facial Recognition JAN - APR 2022

Quest Global, Bangalore - Senior ML Engineer JULY - DEC 2021

  • Developed an unsupervised anomaly detection algorithm to identify irregularities in ambient light sensor data.

Tarsyer, Pune - Computer Vision Engineer SEP 2019 - Mar 2021

  • Implemented and Deployed Vision algorithms for real-time detection and tracking (Kalman Filter) on edge devices.
  • The size of the custom YOLO object detection model was nearly 200kB.
  • Bearing counting model deployed in production: https://youtube.com/shorts/LcIqxHFq1h0
  • Responsible for the scientific and engineering aspects of all machine learning projects.

Swaayatt Robots (Intern) - Reinforcement Learning AUG 2018 - MAY 2019

  • Successfully Tested Deep RL algorithm on self driving (toy) Car. Studied Optical Flow, SLAM etc.
  • Experimented with RL Algorithms like variants of DQN, Vanilla Policy Gradient, PPO, TRPO in OpenAI Gym environment for both Discrete and Continuous action space.

Projects

  • Diffusion Models, Stable-Diffusion, SDXL, ControlNet.
  • Encodec, Coqui TTS, OpenAI Whisper, LLM, nanoGPT, Llama etc.
  • RLHF Training, DPO.
  • Real-time Bearing Counting, Manufacturing Plant Monitoring.
  • Automatic License Plate Recognition, ATM Monitoring using Computer Vision
  • Self driving toy Car

Education

BE, Civil Engineering AUG 2014 - JUNE 2018

Switched to Machine Learning in 2017