Corentin Giaufer--Saubert

- Architected a high-throughput OpenTelemetry ingestion pipeline processing ~1 TB of trace data per day, migrating legacy systems toward a scalable Kafka, ClickHouse, and Quickwit architecture. - Designed ClickHouse storage and query patterns for trace analytics, reducing per-query time from dozens of seconds (and hundreds of GB of RAM) to sub-second filtering and retrieval through partitioning, optimized schemas, and cost-efficient object storage integration. - Developed a custom OpenTelemetry Collector extension with API-key authentication, advanced trace transformation, and a streaming parser for high-performance ingestion. - Solved critical trace consistency issues (multi-day traces breaking partition assumptions), improving data correctness and query reliability. - Led backend modernization across Go and TypeScript services: introduced TDD, stronger domain modeling, and end-to-end test coverage. - Rebuilt the open-source Python and Typescript SDK around OpenTelemetry conventions, adding auto-instrumentation, custom processors, and richer client-side telemetry capture. Stack: Go, ClickHouse, Kafka, AWS, Quickwit, OpenTelemetry, S3, Redis, TypeScript, Python, PostgreSQL, Kubernetes

- Designed event-driven backend services on AWS with clear service boundaries, resilient workers, and asynchronous ingestion pipelines (outbox, ordered queues). - Principal maintainer of a Go chess library — rewrote >50% of the codebase, cutting PGN parsing from 26s to 0.888s (29×) in multithreaded mode and 2× single-thread; replacing the legacy Python parser yielded a 14× boost. - Re-implemented performance-critical workers in Go, trimming CPU usage and noticeably lowering the AWS bill. - Processed 300+ hours of chess video: built OpenCV/FFmpeg pre-filters that sped up an existing CNN board-recognition model ~10×, enabling frame-accurate reverse search of positions. - Ingested thousands of games from Chess.com and Lichess via resilient, idempotent SQS workers; produced detailed game analyses and aggregated player stats. - Built "Lab," a unified store for games, engine evals, and metadata. - TDD with Testcontainers; integrated OpenTelemetry traces for rapid root-cause debugging across services. Stack: Go, AWS (SQS, Lambda, SNS, EKS), PostgreSQL, Python, Redis, OpenCV, FFmpeg, OpenTelemetry

- Designed and shipped a Go-based spaced-repetition flashcard platform end-to-end. - Architected a high-performance REST API in Go implementing a sophisticated spaced-repetition algorithm for personalized learning paths. - Built a responsive UI using HTMX and Tailwind CSS — modern UX without heavy JS. - Established robust cloud infrastructure: Docker, PostgreSQL, Redis caching/sessions. - Implemented full observability stack: Jaeger (distributed tracing), Prometheus (metrics), Metabase (analytics). - Robust test suite with Testcontainers for integration testing; single-binary architecture with distroless Docker images and multi-stage builds.

I contributed to embedded systems development, focusing on protocol implementation and system integration: - Developed and integrated proprietary and Open Protocol communication interfaces in C on a BuildRoot-based embedded system, ensuring robust device-to-device data exchange - Coded a cross-platform desktop application using C++/Qt, enabling real-time WiFi communication with embedded devices for efficient data retrieval, filtering, and analysis - Designed and implemented an intuitive graphical interface for embedded system control, significantly boosting operator workflow efficiency - Created comprehensive technical documentation covering system architecture and protocol specifications