At Nucleovir Therapeutics Pvt. Ltd, I develop protein language model and structure-prediction workflows for sequence design and AI-driven biomolecule discovery. I also build automated pipelines that evaluate, filter, and rank generated protein candidates.
At National Institute of Technology(NIT), Warangal, I deployed lightweight ML and deep-learning models for offline use on Android, Raspberry Pi, and NVIDIA Jetson Nano devices. I reduced model size from 10 MB to 3 MB through quantization, with 230–350 ms end-to-end inference latency on constrained edge hardware.
At Evoastra Ventures Pvt. Ltd., I built fraud-detection pipelines using XGBoost and Random Forest, achieving 92.4% accuracy. I also built a LoRA and DreamBooth fine-tuning pipeline for Stable Diffusion using PyTorch and Hugging Face Diffusers.
