Samuel Braude

About

Hands-on engineer who has taken an LSTM Bitcoin forecasting model from notebook to a production-grade, monitored API (Docker/Gunicorn/Nginx, Prometheus/Grafana, blue-green rollback) delivering 99.9% availability and ~110–120ms p95 latency. Also built an RFID self-checkout prototype spanning Raspberry Pi + firmware + networking, using deep instrumentation to eliminate double-charges/timeouts (<0.1%) and reduce checkout time ~20% through idempotency, debounce logic, and hardware fixes.

Experience

Education

Key Strengths

• Productionizing ML models into reliable APIs (Flask/Docker/Gunicorn) with deterministic builds
• Reliability engineering with SLOs, Prometheus/Grafana alerting, health checks, and blue-green rollback
• Performance optimization for inference (model preloading, vectorization, background updates via Redis worker) achieving p95 ~110–120ms
• End-to-end debugging across hardware/network/software using logic analyzer + tcpdump/iperf; reduced duplicate/timeout rate from ~3–5% to <0.1%
• Designing idempotent transaction flows (idempotency keys, retries/timeouts, circuit breakers) for real-world edge cases
• Customer/operator-facing iteration: on-site shadowing, rapid PoCs, daily iteration, training + SOPs; improved checkout time ~20%
• Building customer-specific Python services from fuzzy specs (FastAPI/Celery) cutting report time from hours to minutes
• IoT/edge-to-cloud and on-prem Linux networking fundamentals (DHCP/VLANs, MQTT/HTTP) with hands-on Wi-Fi diagnostics