Software Engineer · Backend, AI Tooling & Edge Systems
Software engineer with five years of experience building production systems across the stack: backend services and APIs, data models, frontends, and cloud infrastructure. My focus is reliability and observability, and my recent work extends into AI tooling and edge systems.
▶ Watch the 85-second demo: an ESP32 continues collecting IMU data through a 30-second broker outage. Telemetry lost: 0 records.
ESP32 · C++ · FreeRTOS · MQTT · Python · TypeScript · InfluxDB · Grafana · MCP
IMU → ESP32/FreeRTOS → PSRAM buffer → MQTT → Raspberry Pi gateway → InfluxDB → Grafana
A fault-tolerant edge telemetry system for retrofitting legacy industrial equipment with sensing and remote monitoring. It maintains telemetry through network outages using local buffering, controlled replay, and acknowledgment after gateway persistence.
- PSRAM-backed ring buffer with drop-oldest eviction under pressure
- Bounded edge buffering with controlled replay after reconnect
- State machine recovery: NORMAL → BUFFERING → SYNCING,
with
boot_idand sequence tracking across disconnects - TypeScript MCP server for querying live telemetry through an LLM
In progress: end-to-end backpressure propagation from cloud ingest to edge.
Python · LLM APIs · Docker · MCP
A provenance-aware knowledge pipeline that validates AI-generated content before it's published.
- Deterministic validation with source and content-hash checks, conflict detection, and human approve/reject/defer workflows
- Create-only publisher with transaction journaling, crash recovery, and guarded rollback
- Network-isolated Docker validation sandbox and 300+ tests
I design around failure modes first: unreliable networks, bounded resources, duplicate delivery, race conditions, and partial recovery. Backpressure, idempotency, fault isolation, and explicit state management apply whether the system is a high-traffic web service, an AI pipeline, or a microcontroller with tight memory limits.