🔭 I’m currently working on
Low-latency LLM infrastructure at Amazon Web Services, building systems with SigV4 + LDAP authentication for production clients. I also optimize distributed observability pipelines across 150+ AWS services using OAuth-backed sync platforms, ECS Fargate, and Docker.
👯 I’m looking to collaborate on
Projects involving HPC, scalable machine learning systems, or efficient inference backends—especially ones pushing token throughput, optimizing distributed performance, or innovating in edge deployment.
🤝 I’m looking for help with
Advanced GPU kernel-level tuning and low-latency optimization techniques for multi-modal LLMs. Also interested in learning more about quant trading infra or novel compression algorithms for AI/ML inference.
🌱 I’m currently learning
Deep dive into distributed training frameworks, especially gradient checkpointing and tensor parallelism for multimodal models. Also brushing up on real-time streaming data systems and reinforcement learning for ops optimization.
💬 Ask me about
Achieving 120 tokens/sec inference on Llama-3 using vLLM
Cutting runtime by 45 seconds at NIST with multiprocessing and Numba
Engineering sub-5 ms auth for AWS clients with Coral + Guice
Developing GraphQL backends and Kafka pipelines at Clinia
⚡ Fun fact
I once boosted image processing throughput by 720% using a 72-core EC2 HPC setup—on my own—and I still have the benchmark logs to prove it.
