Pratik Doshi — AI Researcher

Hello, and welcome to my LinkedIn profile! I am passionate about designing and deploying AI systems that tackle real-world challenges with speed and efficiency. Applied AI & ML Systems Expertise: - Fine-tuned VLMs, LLMs, and CV models for object detection and segmentation on domain-specific tasks. - Optimized inference serving using NVIDIA TensorRT, ComfyUI, and vLLM to achieve low latency and high throughput. - Built scalable PyTorch-based inference services leveraging Flash Attention and Speculative Decoding. I earned my MS from the University of California, Santa Cruz, focusing on Mechanistic Interpretability, AI Safety, and ML systems. My research centered on reverse-engineering neural networks to improve transparency and robustness, and I’ve trained deep neural networks — LMs, VLMs, and classification systems—on large datasets in GPU-accelerated clusters. As an AI/ML Consultant, I've built several AI-powered applications and led several end-to-end ML projects — from data preparation and feature engineering to model training and continuous evaluation. Before that, as a Software Engineer, I architected backend systems that ingested, stored, and analyzed massive streams of financial market data, ensuring reliability and performance at scale. I enjoy connecting with others who are excited about Applied AI and ML Systems. If you’re interested in chatting about inference optimization, model explainability, or any AI-driven innovation, feel free to reach out at pratikdoshi99@gmail.com.

Stackforce AI infers this person is a Machine Learning Engineer with expertise in AI and Fintech.

Location: Mountain View, California, United States

Experience: 2 yrs 3 mos