Warre Snaet
About
I’m a backend and applied AI engineer from Belgium. I graduated with distinction in AI Engineering at Howest.
My strongest work is in Python: APIs, data pipelines, computer vision, OCR, and applied machine learning, including reinforcement learning and generative AI. I also work with AWS and Azure, Terraform, containers, Kubernetes, and CI/CD, with a particular interest in agent infrastructure and MLOps.
I’m open to junior cloud, backend, software, data, and applied AI engineering roles.
Experience
Medical imaging at 2Ai IPCAAI & software engineering intern
International internship using real clinical imaging data. I built a Python DICOM/OCR pipeline, validation and review tools, resumable batch processing, ECG extraction, and a source-video matcher. Metadata-guided alignment reduced one representative run from about 23 hours to 21 minutes.
Apolloon 24-hour run systemSoftware engineer, paid client work
I built local event infrastructure where every Electron host owns a writable SQLite database. Hosts pair once, rediscover peers through signed UDP on the LAN, and reconcile only missing operation deltas over authenticated HTTP.
Selected work
AWS Fargate & Vault workload identity
A tested Terraform architecture for private ECS Fargate workloads behind an ALB, with two-AZ networking, IAM-based Vault authentication, CloudWatch observability, autoscaling, health checks, and deployment safeguards.
Financial AI agent
A six-service application with FastAPI streaming, MCP tool execution, RAG, PostgreSQL, React, ChromaDB, and local Ollama inference. In this two-person academic project, I owned the backend architecture, orchestration, persistence, containers, tests, and most frontend integration.
Azure ML lifecycle
Individual MLOps coursework covering Azure ML compute, parallel data preparation, model registration, MLflow, FastAPI, Docker, Kubernetes manifests, and GitHub Actions.
Semi-supervised learning in Rust
Bachelor research into data-efficient plant-disease classification on edge devices. The saved SSL checkpoint reached 94.90% held-out accuracy and shipped as an approximately 26 MB offline application tested on iPhone.
Event chatbot for XPO Group
Academic team project for an external client: a full-stack RAG system using Azure OpenAI, Cosmos DB, .NET, Next.js, and Python ingestion.
Dataset query system
A Python client-server analysis application with authentication, moderation, statistics, broadcasts, and a PySide6 interface.
Open source
Athas code editorMaintainer & contributor
I contribute across its React/TypeScript and Tauri/Rust codebase. Merged work includes C# language support, image diffing, Linux UI improvements, dynamic proxy ports, resource- and memory-leak fixes, and an LSP race-condition fix.
Writing
How I built a 26 MB offline AI with Rust and Burn
The research and implementation behind my bachelor project, from semi-supervised learning to running the model on a phone.