Manish Pandey

About Manish Pandey

I am a dedicated Machine Learning Engineer and co-founder of RoentGen Health, focusing…

Experience

1. Machine Learning Engineer

   KPMG India

   Jan 2023 — Present · IN

   Staying updated on LLM Research and Development, and developed LLM model to increase Context Learning (ICL) from 32K to 1M tokens with Linear Attention, inspired by Linear Transformer’s Encoder-only Architecture. I worked on a parallel graph attention neural network for the multi-type extraction of knowledge graph embeddings. Achieved features of diverse graph types DAGs, Weighted Graphs, and Hierarchical Temporal Graphs for enhanced analysis and interpretation of complex medical imaging data. Managed Distributed Training GPUs Data Parallelism for Training LLM. Achieved 30% reduction in training time and 25% improvement in inference speed. Provided Consultation and Advice on building a Mixed-Agent Framework. Integrated Autonomous Agents with Retriever Augmented Generation Pipeline. building chat-based AI assistants through personalized solutions Fine-Tuning and Quantization. Handling the API, LLM, model deployment, evaluation frameworks, end-to-end ML pipeline, and unit testing. Experimented with a couple of open LLMs and sentence embedding models during the process like LLAMA, alpaca, Mistral-7b, DistilBERT, all-MiniLM-L6-v2, etc.