Ilmari Vahteristo — Computer vision and machine learning engineer

I build perception systems that run in the real world: cameras, LiDAR, radar and lasers on edge hardware, in factories and heavy industry. At a seven-person startup I worked directly with customers and built four systems that are now deployed.

Based in Helsinki and open to new opportunities — happy to talk over a call or coffee.

email meCV (PDF)LinkedInGitHub

Experience

  • 5/2025 – 8/2026
    Software Engineer, ML / CV · VirelabsA team of seven. Worked directly with customers; all four systems are deployed.
    • Multi-camera tracking through occlusions — 5 cm at 6 m
    • 3D reconstruction of moving vehicles, line cameras + LiDAR + radar — 4 cm
    • Veneer quality control by laser triangulation — 0.5 mm from over 1 m
    • OCR on industrial packaging, with majority voting — 99.9 %
  • 4/2024 – 8/2024
    Research Assistant · LUT UniversityDeep learning for limited-angle CT. First author of the paper; presented it at SCIA 2025.
  • 1/2024 – 4/2025
    Freelance data scientistPredicted restaurant customer volume from the menu, to cut food waste.

Leadership

  • 1/2021 – 12/2022
    Project Manager · Lateksii, LUT student guildLed the organizing team and volunteers; events of up to 900 people.
  • 11/2020 – 7/2021
    Peacekeeper · UNIFIL, LebanonSquad co-lead; promoted to part-time squad lead, then team co-lead.

Publications

  • 2026
    New spherical codes by tuning an annealing solverarXivcoderecordsPreprint, sole author. Holds over 3,000 of the current records on spherical-codes.org.
  • 2025
    Data-efficient limited-angle CT using deep priors and regularizationSpringerarXivSCIA 2025 (Springer LNCS). First author and presenter.
  • 2025
    Stationary LiDAR, moving target: pure LiDAR solution for motion estimation and 3D reconstructionpdfMaster's thesis, LUT University, with Virelabs.
  • 2023
    Man vs machine: beating humans in a multiplayer card game without lookaheadpdfBachelor's thesis, LUT University.

Projects

  • Moska
    Card-game AI that beats top human playersmoska.euNow also a benchmark of how LLMs play a game they have never seen.
  • PolyGate
    MCP server for LLM agents on prediction marketsgithub

Competitions

  • 2025
    Junction — won the ElevenLabs challengedemo
  • 2024
    Junction — top 5 % in peer reviewdemo
  • 2024
    Kaggle LLM 20 Questions — gold medal, 7th of 800
  • 2023
    Visma coding competition — 1st of 20 teams
  • 2022
    IndySCC supercomputing — 3rd of 10 universities

Education

  • 2025
    MSc (Tech), Computer Vision · LUT UniversityGPA 4.2 / 5. Exchange at IST Lisbon.
  • now
    MSc, Robotics and Automation · Aalto UniversityPart-time.

Skills

  • Sensors
    Cameras · LiDAR · radar · laser triangulation · IMU
  • Methods
    Deep learning · 3D vision · tracking · sensor fusion · optimization
  • Tools
    Python · PyTorch · TensorFlow · ONNX · OpenCV · C · Docker · Git · Linux
  • Compute
    Edge devices · PLC · parallel and supercomputing
  • Languages
    Finnish (native) · English (fluent)

Contact

i.vahteristo@gmail.comlinkedin.com/in/ilmarivgithub.com/ilmari99Google ScholarCV (PDF)

Visits are counted with Umami — no cookies, nothing personal.