AI/ML engineer in Oulu, Finland. MSc in Computer Science and Engineering, University of Oulu. Founder of an early-stage robotics-AI startup.
-
Building a federated fleet-learning layer for robots: fleets improve from each other's experience without sending raw data to the cloud.
-
MSc thesis: a multi-agent LLM marketing platform with a built-in safety and governance layer. It beat a rule-based baseline by 89% in simulation, was graded 5/5, and was funded by a 20,000 euro research grant.
-
Before that, two years at Cowlar (Y Combinator W17) on embedded and machine-vision systems, including a machine-vision fiber-aligner that took output from 15 to 120 units an hour.
-
agentic-marketing-platform-public: de-identified public release of the thesis platform, with LangGraph agents over a SimPy market simulation, pgvector RAG, contextual bandits gated by off-policy evaluation, and human-in-the-loop governance.
-
turtlebot4-people-avoidance: people-aware TurtleBot4 navigation in ROS 2, from LiDAR leg detection through multi-target Kalman tracking to a Control Barrier Function safety filter, with a live web dashboard.
-
BloomOulu: platform for the University of Oulu Botanical Garden, a Next.js + NestJS monorepo with a local RAG plant guide, donations, QR labels, and a kiosk (live demo).
-
looking-glass: natural-language video search over camera footage, fully local, built on YOLO-World, SigLIP, Qdrant, Ollama, FastAPI, and React.
-
SmartConveyor-ICS: conveyor industrial-control simulation on Kubernetes, with leader-gated gRPC servers, Kafka telemetry into InfluxDB, and CrewAI agents on Ollama.
-
EyeAI-Retinopathy-Detection: diabetic retinopathy grading on DeepDRiD with PyTorch, using transfer-learning backbones, bagging/boosting/stacking ensembles, and GradCAM explanations.
-
Languages: Python, C/C++, TypeScript
-
ML: PyTorch, LangGraph, LangChain, RAG with pgvector and Qdrant, Ollama, contextual bandits with off-policy evaluation
-
Backend: FastAPI, NestJS, Next.js, PostgreSQL, Redis, Kafka, gRPC, MQTT, InfluxDB
-
Robotics and edge: ROS 2, TurtleBot4, MicroPython on the Raspberry Pi Pico W
-
Infra: Docker, Kubernetes, Prometheus + Grafana, GitHub Actions