At Deutsches Klimarechnenzentrum(DKRZ), I optimized GPU-direct MPI communication and achieved up to 20% performance gains with 1000+ GPUs. I also developed CPU/GPU parallel computations and optimized parallel I/O on Lustre.
At Verda Oy, I worked on GPU cluster benchmarking, validation, and reliability engineering. I developed benchmarks and Grafana and Prometheus visualizations, helping move Verda’s ClusterMax rank from 41st to 13th.
In my current Agentic AI Hackathon & OpenSource projects, I won first place in the Nebius and NVIDIA Hackathon’s Healthcare and Life Sciences track with an LLM program that searches and ranks research papers. I also developed an agentic AI infrastructure diagnosis tool tested on real HPC telemetry.

