Systems engineer in New York City focused on ML infrastructure, low-latency C++, deep learning systems, and distributed infrastructure.
I have spent nine years building startup systems across language-model evaluation, real-time networks, cloud infrastructure, developer tooling, and product platforms. My current work is centered on understanding and building the machinery behind modern deep learning systems.
A code-first video series that rebuilds deep learning from first principles in modern C++.
The implementation currently includes tensor storage and indexing, reductions, elementwise operations, matrix multiplication, broadcasting, mean squared error, computation graphs, and scalar reverse-mode automatic differentiation. The project will continue toward tensor autograd, training loops, neural networks, transformers, performance engineering, GPU programming, and custom CUDA kernels.
An encoder-decoder transformer trained to generate tactical commentary for chess moves. The project covers data mining, quality analysis, custom tokenization, model training, and post-training evaluation with PyTorch and Hugging Face.
Source and notebooks · Video walkthrough
At Haize Labs, I built the core test harness used for red-team and functional testing across hundreds of thousands of daily language-model conversations, their evaluations, and their taxonomies. I also led a directed-graph interface for maintaining large suites of functional model tests.
I authored a caching RPC load balancer in Rust and integrated it into production infrastructure. It served roughly 1,000 to 2,000 requests per second on 0.5 vCPU and 250 MB of memory, cut upstream API costs by about 50%, and saved more than $20,000 per month.
The gateway supports HTTP and WebSocket traffic, caching, health monitoring, load balancing, Docker, Kubernetes, and configurable observability.
At Here Not There, I led cloud infrastructure for a decentralized, real-time, end-to-end encrypted messaging network. The work included Kubernetes, observability for externally operated nodes, reproducible sandbox environments, protocol infrastructure, and a migration from AWS ECS to GCP Kubernetes.
I joined Matter as its first employee and led architecture across databases, caching, queues, serverless workloads, GraphQL APIs, testing, CI, and frontend infrastructure. I also built recommendation systems and an internal A/B testing platform.
Earlier open-source work includes Uniform GraphQL, a code-first type-safe GraphQL framework, and VulcanJS CLI, a community-adopted scaffolding tool distributed through npm.
- Languages: C++, Rust, Python, TypeScript
- Machine learning: PyTorch, transformers, model evaluation, data pipelines, automatic differentiation
- Systems: distributed systems, caching, load balancing, observability, profiling, resource efficiency
- Infrastructure: Kubernetes, Terraform, Docker, AWS, GCP, Cloudflare, Datadog
- Currently studying: numerical computing, deep learning internals, CUDA, and custom GPU kernels
- BSE in Computer Science, University of Pennsylvania
- 0xMacro Secure Smart Contract Engineering Fellowship, honors tier
- Product Hunt Golden Kitty Award for Social Impact with Matter
- Co-inventor on a pending patent for automated performance feedback