I occasionally take on short-term contracts with startups and companies building performance-critical software and large-scale distributed systems. I build prototypes and initial products, optimize low-level execution paths, and solve difficult latency, concurrency, and reliability problems.
Discuss a project by emailCPU profiling, memory allocation, cache locality, data layout, synchronization, and network I/O. I trace execution paths, identify contention and unnecessary work, and optimize the parts of a system that determine throughput and latency.
From nanosecond-scale CPU and cache behavior to microsecond-scale execution paths and millisecond-scale service budgets. I measure p50, p95, and p99 latency under representative load, isolate queueing and scheduling delays, and set targets for each stage of the critical path.
Partitioning, replication, consistency, coordination, and load distribution. I design and debug systems that need to operate through partial failures, traffic spikes, and uneven load, with explicit decisions about data ownership, capacity, backpressure, and failure isolation.
Detect recoverable failures, preserve work, and safely retry or replay it within the operation’s time budget. I use idempotency, durable state, and deduplication to protect against duplicate effects, and reconciliation to resolve uncertain outcomes. Bounded retries and backpressure keep recovery from amplifying overload. The goal is to resolve failures before users encounter them.
A short contract starts with a concrete question and an agreed scope. Examples of the work we can define together:
Share the system or product you’re building, the target workload, and the latency or reliability constraint. Profiles, traces, and reproducible failures are useful starting points when available.
david@permissionless-labs.com