Data Science Intern | Microsoft IDC, Hyderabad
Prototyped a ReAct-based anomaly-detection pipeline for Windows telemetry, reducing runtime from 30 minutes
on a 64 GB CPU to a projected 1 minute on a laptop GPU. Designed a privacy-preserving, on-device inference
pipeline with Llama 3.1–8B, with potential to cut diagnostic cloud calls by 90% and save 1,200+ support hours
per 1,000 tickets.
CTO & Co-Founder | NeuralThread.AI, Mumbai
Led development of multimodal GenAI pipelines using Stable Diffusion and ESRGAN, reducing apparel
design effort by over 80%. Directed AI apparel visualization with StyleGAN, enabling scalable mock-ups
with efficiency gains of 70–90%.