Backend lead on Ford Pro's EV charging platform — 8,200 connected chargers, 6M+ telemetry messages/day through Kafka at 99.9% data integrity. I architected the OCPP gateway that replaced the legacy monolith: migrated the full fleet, decommissioned ~2,150 servers (~$9k/mo), and cut Datadog spend 60% by fixing what we were logging rather than negotiating the bill. I own L3 for these services, so I've spent real time on what happens when firmware, flaky cell links, and a protocol spec that vendors each implement creatively all meet in production.
The most interesting thing I've fixed recently: a Node.js Kafka consumer doing fire-and-forget processing with no concurrency limit. Under downstream degradation, retry timers filled the microtask queue faster than GC could reclaim it and the process OOMed, taking messages with it. Redesigned it around circuit breakers, per-region bulkhead isolation, Redis-backed per-partition backpressure queues preserving in-order delivery, and Kafka offset commits gated on replica confirmation - so it degrades predictably and self-heals instead of paging someone.
Side project: PhotonicOps, a Go telemetry ingestion engine for silicon photonic biosensors - 10kHz over client-streaming gRPC, sync.Pool hot path, lock-free ring buffer, fully air-gapped with zero cloud API calls. Phase 1 of 5 is built; the rest is ADRs, and the README says which is which.
Looking for backend or platform work on connected devices, telemetry, or energy systems. Go preferred. Set up for contractor/USD invoicing.
Remote: Yes (only)
Willing to relocate: Yes - EU/UK for the right team
Technologies: Go, Kafka, gRPC/Protobuf, OCPP 1.6/2.0.1, PostgreSQL, MongoDB, Redis, Docker, AWS, NestJS/Node, TypeScript, Python, Datadog
Résumé/CV: https://clint-mathews.github.io/CLINT-MATHEWS.pdf
Email: mathewsclint28@gmail.com
Backend lead on Ford Pro's EV charging platform — 8,200 connected chargers, 6M+ telemetry messages/day through Kafka at 99.9% data integrity. I architected the OCPP gateway that replaced the legacy monolith: migrated the full fleet, decommissioned ~2,150 servers (~$9k/mo), and cut Datadog spend 60% by fixing what we were logging rather than negotiating the bill. I own L3 for these services, so I've spent real time on what happens when firmware, flaky cell links, and a protocol spec that vendors each implement creatively all meet in production.
The most interesting thing I've fixed recently: a Node.js Kafka consumer doing fire-and-forget processing with no concurrency limit. Under downstream degradation, retry timers filled the microtask queue faster than GC could reclaim it and the process OOMed, taking messages with it. Redesigned it around circuit breakers, per-region bulkhead isolation, Redis-backed per-partition backpressure queues preserving in-order delivery, and Kafka offset commits gated on replica confirmation - so it degrades predictably and self-heals instead of paging someone.
Side project: PhotonicOps, a Go telemetry ingestion engine for silicon photonic biosensors - 10kHz over client-streaming gRPC, sync.Pool hot path, lock-free ring buffer, fully air-gapped with zero cloud API calls. Phase 1 of 5 is built; the rest is ADRs, and the README says which is which.
Looking for backend or platform work on connected devices, telemetry, or energy systems. Go preferred. Set up for contractor/USD invoicing.