The systems layer under machine learning: training libraries, checkpoints, agent orchestration, LLM observability and developer tooling.
I've been building software since 2018, most of it in Rust: async services, FFI, memory allocators and protocol code that has to stay correct under load.
Today I work where that background matters most in AI: the infrastructure that models and agents run on. That means numerical correctness in training code, checkpoints that survive a crash, evaluation scores you can trust, and traces that tell you why an agent did what it did.
Models are only as reliable as the systems that train, evaluate and serve them.
| Area | What I focus on |
|---|---|
| Training library internals | Optimizer math (RMSprop variance, epsilon placement), weight initialization (Kaiming/Xavier), RNN input validation, numerically stable edge cases |
| Checkpointing & data integrity | Atomic checkpoint writes, batched tensor serialization, verified dataset downloads |
| Agent orchestration & safety | Lock lifecycle in task orchestrators, eval-dimension scoring, per-agent event types, scope enforcement |
| LLM observability | OpenTelemetry spans for agent and tool events, error capture in LangChain tracing, token-usage parsing, AI memory and knowledge planes |
| Grafana & monitoring | ECharts plugin hardening, template variable regex, Pyroscope datasource fixes |
| Forecasting & anomaly detection | Exogenous-feature support in time-series foundation-model pipelines |
| Systems foundations | Tokio, cancellation safety, backpressure, zero-copy I/O, FFI and unsafe audits |
All links go to my own pull requests. 55+ merged across 30+ repositories.
| Project | Contribution | Status |
|---|---|---|
| grafana/dataViz: Grafana ECharts plugin monorepo | Guard non-string config.src in getLayerSrc · Replace untracked plan file references | in review |
| grafana-pyroscope-datasource | Fix multi-value template variable regex | in review |
| getsentry/sentry-go: Sentry SDK for Go | Add WithMechanism capture option | in review |
| lemma: LLM tracing | Record handled LangChain tool errors as span errors | in review |
| oxtrace: LLM usage tracing | Parse usage from gzip-encoded responses | in review |
| foundationforecast: time-series foundation models | Forward futr_exog_list in detect_anomalies | in review |
| Project | Contribution | Status |
|---|---|---|
| eza-community/eza: modern ls replacement | Add EZA_SIZE_STYLE env var | in review |
| beancount-language-server: LSP for Beancount | Advertise positionEncoding capability (LSP 3.17) · Fix exit code on unclean shutdown | merged |
| tree-sitter-beancount: grammar | Add injections.scm for BQL query highlighting | in review |
| metal2vk: Metal-to-Vulkan translation | Key pipeline variants by library/name/constants · Reject inconsistent push/descriptor metadata | merged |
| smolvm: lightweight VM manager | Let a disk volume set its directory's owner and mode | in review |
| mimas: GML parser in Rust | Accept loop, while, for, if and match as binary operands | merged |
Own repositories: smalloc (memory allocator for Solana programs), Dbgeng-sys (Rust FFI bindings for the Windows debugger engine), Solana-Credit-Score, AsyncCSVX.
01. Measure, then optimize. Throughput, tail latency and allocations, not guesses.
02. Make failure boring. Writes are atomic, inputs are validated at the boundary, and errors say what to do next.
03. Trust numbers only when they're tested. An eval score or a loss curve is only as good as the code that computes it.
04. Small, reviewable changes. Every PR above is scoped to one problem and easy to review.
Open to remote roles and contracts with AI/ML startups: ML infrastructure, inference, evaluation and Rust performance work.