🔧 AI/ML Engineer building multi-agent systems and GPU inference optimization
🎤 Demonstrated Helion GPU kernels at NVIDIA GTC 2026
🏆 GPU MODE Helion kernel contest (B200): 1st & 2nd, plus 4th & 6th
🔬 Merged PRs in Google Tunix, LangChain, and Ivy
📄 Co-author, SemEval-2025 Task 3 Best System Paper (hallucination detection)
🎓 MS in AI (NLP), UC Santa Cruz (3.9 GPA)
✉️ Reach me at karthikraja2k1@gmail.com
Know more about me at ProfilePage
| Project | What it is | Page |
|---|---|---|
| Morph | Self-evolving harness: LLM agents compose and evolve skills for robot fleets whose bodies change (MuJoCo) | view |
| Smart Router | Multi-modal LLM router — picks the cheapest capable model in ~15 ms with guardrails and budgets | view |
| kernel-dev | Kernels across MLX, CUDA, Triton and JAX/TPU with roofline analysis | view |
| helion_optimizations | Helion DSL kernels for the GPU MODE Helion Kernel Challenge (B200) | view |
| StaffingGym | RL environment: an LLM agent runs a staffing agency over a simulated 52-week year (OpenEnv Hackathon) | view |
| TeamHQ | Engineering org command center for autonomous code change (Nozomio Hackathon) | view |
| LLM Evals | Multi-model evaluation with RAG, voice, tracing and regression detection | view |
| Gui_Agent | GUI agent for autonomous web-task completion (capstone research) | — |
| event-capture | Captures browser event trajectories to drive and evaluate web agents | — |
| tunix | Google Tunix (JAX post-training) — merged PR #1552 | — |
| Language |      |
| ML Framework |        |
| Web Development |    |
| Databases |   |
| OS |   |
