Prakhar Saxena
Model Training | Generative AI | Startups
Model Training | Generative AI | Startups
Hi there! I'm a AI researcher and engineer with roots in Electrical and Electronics Engineering.
As a researcher, my work involves emotion classifiation and LLM alignment.
As an engineer, I help business and people with generative AI (Text, Image and Audio).
Outside my professional life, I find joy in nature and draw inspiration from both Eastern and Western philosophy..
Led the development and deployment of HighEQ (Conversationalist) LLM, overseeing continued pretraining, supervised fine-tuning (SFT), reinforcement learning (RL), and deployment on Inf2 infrastructure.
Reseach on data deduplication for LLM training at LegalForce, comparing NCD and Edit Similarity on Wiki40B. Training GPT-2 on cleaned data led to improved perplexity over baseline.
Built a BERT-based sentiment analysis model using PyTorch and Hugging Face for Insight Tech, a Tokyo-based company that analyzes online reviews. Improved overall F1 score by up to 20% across sentiment classes.
.
Feel free to reach out.