2026 CS grad looking for my first full-time role. I build things that
have to actually work, and I spend most of my time on the part where
they don't.
KARMA — KARMA (github.com/ShreySharma07/Multi-Agentic-CUDA-optimization): a self-improving multi-agent system that writes and tunes GPU code. Five specialized agents profile, plan, code, test and reflect, with a vector memory that carries lessons from one kernel to the next. github.com/ShreySharma07/Multi-Agentic-CUDA-optimization
Repliq — a computer-use agent that turns one narrated screen recording
into a self-verifying execution plan. Runs a 20-step Salesforce workflow
end to end. Same lesson, different shape: it used to report every run as
a success while records silently failed to save. Fixed by giving each
primitive a postcondition contract checked against real application
state, not against the click returning.
github.com/ShreySharma07/salesforce-agent
Currently working on developing novel technique to post-tune Gemma 4 31B model for swe tasks.
Remote: Yes
Willing to relocate: Yes — anywhere. India, EU, US, UK.
Technologies: Python, PyTorch, Redis LLM inference, LLM fine-tuning (Qlora, peft), Google ADK, Agentic systems, FastAPI, Playwright, Docker, SQL, AWS, Tensorflow, MLflow, MCP, RAG, Neo4j, Git/Github, CI/CD, Typescript, API, scikit-learn, Vector Databases, HuggingFace.
Resume - https://drive.google.com/file/d/1qUxHkYu8dx866aGJy4-GLejJui7...
Email: [email protected]
---
2026 CS grad looking for my first full-time role. I build things that have to actually work, and I spend most of my time on the part where they don't.
KARMA — KARMA (github.com/ShreySharma07/Multi-Agentic-CUDA-optimization): a self-improving multi-agent system that writes and tunes GPU code. Five specialized agents profile, plan, code, test and reflect, with a vector memory that carries lessons from one kernel to the next. github.com/ShreySharma07/Multi-Agentic-CUDA-optimization
Repliq — a computer-use agent that turns one narrated screen recording into a self-verifying execution plan. Runs a 20-step Salesforce workflow end to end. Same lesson, different shape: it used to report every run as a success while records silently failed to save. Fixed by giving each primitive a postcondition contract checked against real application state, not against the click returning. github.com/ShreySharma07/salesforce-agent
Currently working on developing novel technique to post-tune Gemma 4 31B model for swe tasks.