Recent CS grad building AI systems from the ground up. Built a multi-agent system that optimizes CUDA kernels using real NVIDIA Nsight Compute profiling data fed into LLM prompts (1.08x–1.31x speedup on KernelBench Level 1, 0% compile failure). Also built an autonomous AI agent platform (Repliq) that converts screen recordings with voice into executable workflows using ReAct loops, tool-calling, and zero-trust sandboxed execution, shipped it to salesforce employee for automating redundant task in salesforce. Looking for AI/ML engineering roles — happy to talk about either project in depth.
Willing to relocate: Yes (US, Europe)
Technologies: Python, PyTorch, FastAPI, Docker, LLM agents (ReAct, MCP tool-calling, A2A protocol), RAG, async/await, PostgreSQL, AWS (SageMaker, Lambda, Step Functions, DynamoDB), speect-to-text, DeepLearning, QLora, Peft, sklearn
Résumé/CV: https://docs.google.com/document/d/153Tu_DBgt-OGOw1i607_Yj40...
Email: shrey7shrey@gmail.com
Recent CS grad building AI systems from the ground up. Built a multi-agent system that optimizes CUDA kernels using real NVIDIA Nsight Compute profiling data fed into LLM prompts (1.08x–1.31x speedup on KernelBench Level 1, 0% compile failure). Also built an autonomous AI agent platform (Repliq) that converts screen recordings with voice into executable workflows using ReAct loops, tool-calling, and zero-trust sandboxed execution, shipped it to salesforce employee for automating redundant task in salesforce. Looking for AI/ML engineering roles — happy to talk about either project in depth.