Ville Lähteenlahti
Machine Learning Scientist · PhD in Physics
About Me
I work as a Senior AI Engineer at Teamit, where I build AI that understands and reasons about CAD objects, and evaluate model performance. Before that, I spent five years owning machine learning end to end at Linear Oy, with generative AI in production since 2021. I am hands-on across the whole stack: software architecture, modeling and evaluation, data pipelines, APIs, infrastructure, and deployment. I hold a PhD in neuromorphic memristor devices (hardware for non-conventional computing), with cleanroom fabrication, modeling, and device characterization in a low-temperature physics laboratory. My research experience translates into solid problem-solving skills, precise reporting, and the ability to quickly familiarize myself with complicated systems.
Work Experience
Teamit
Senior AI Engineer • Nov 2026 - Present
- Build AI that understands and reasons about CAD objects, and evaluate model performance.
Linear Oy
Machine Learning Scientist, ML lead • Jul 2021 - Oct 2026
- Owned all machine learning at Linear, running projects from problem definition to production operation.
- Shipped an LLM-based property-description service to production in 2021, a year before ChatGPT.
- Led Linear’s Business Finland-funded GenAI research project on multimodal property similarity.
- Built geospatial prediction models, interactive chat agents with RAG, unstructured-to-structured data processing, and full-system LLM translation.
- Led a data-migration project from multiple real-estate systems into Linear’s platform, moving the data of over 100 companies without incident.
- Designed, implemented and deployed the systems (Azure, PyTorch, XGBoost, Python, PostgreSQL).
University of Turku, Wihuri Physical Laboratory
Doctoral Researcher (Physics, Neuromorphic Devices) • Feb 2018 - Oct 2022
- Fabricated functioning memristor devices (hardware synapses for neuromorphic computing), which established the laboratory’s memristor research line.
- Designed measurement instrumentation and wrote the control and analysis software (Python: PyTorch, scikit-learn, numba, pandas).
- Followed the state of the art of neuromorphic devices; worked with neural networks and device simulations.
- Seven peer-reviewed publications, two as first author; co-inventor of a patented memristor material that won a national inventors’ competition (2024).
- Research led to a follow-up Business Finland project at the university (over 600 000 € funding).
- Worked in a low-temperature physics laboratory: SQUID and PPMS magnetometry for thin-film characterization, and cryogenic handling.
- Teaching assistant: Thermodynamics, Mechanics, Electromagnetism and Atomic & molecular physics.
Nordic ID Oy
Cloud Software Engineer • Jan 2021 - Jul 2021
- Deployed services, configured networks and set up managed databases on Amazon AWS.
- Designed and implemented the access-control system for an autonomous store (PostgreSQL, Redis, FastAPI, OpenAPI).
- B2B integration and data processing related to payment and inventory handling.
Earlier experience
Research, internship and electrical work • 2011 - 2018
- Project Researcher, University of Turku, Wihuri Physical Laboratory (2017–2018): researcher and IT administrator of the laboratory; interfaced and programmed scientific instruments; designed an electrical measurement system for low-temperature and memristor research.
- Research Assistant, University of Turku, Department of Physics and Astronomy (2016–2017): programmed scientific instruments and designed measurements; fabricated nanoscale PrCaMnO thin films for colossal magnetophotoresistance studies; low-temperature physics and cryogenics.
- Software Engineer Intern, Suomen Prosessiautomaatioasennus Oy (2012–2013): automated the scraping of product information for an online store with Python.
- Electrical Assembler, UTU Oy (2011–2012): assembled and wired electrical distribution boards and made sure they passed quality control.
Education
University of Turku, Wihuri Physical Laboratory
PhD in Physics (solid state physics, neuromorphic devices) • 2018 - 2022
- Thesis subject: Memristor devices based on low-bandwidth manganites.
- Machine learning theory, simulations, instrumentation, experiment design and software development.
- Collaboration, presentation and analytical skills.
University of Turku, Department of Future Technologies
Courses in Software Development, Data Analytics and Machine Learning • 2018 - 2020
- Completed the MSc curriculum of computer science alongside the PhD, with emphasis on machine learning.
- Attended summer schools and workshops in high performance computing (HPC).
- The studies included deep learning, distributed systems, algorithm design, SWE best practices and game development.
University of Turku, Department of Physics and Astronomy
MSc in Physics • 2013 - 2017
- BSc and MSc both with the highest grade.
- Minors in computer science and mathematics, including advanced calculus and algorithmic mathematics.
- MSc thesis: Manufacturing and characterisation of PrCaMnO-based resistive memory.
- BSc thesis: Manufacturing, materials and theory of oxide-based memristors.
Ulvilan Lukio, Sataedu Ulvila
Finnish Matriculation Examination and Vocational Qualification, Electrical and Automation Engineering • 2010 - 2013
- Learned the electrician’s trade and completed the Finnish matriculation examination.
- Physics: Laudatur, Advanced math: Eximia cum laude approbatur.
- Diploma work on PLC programming and interfacing of sensors. Included CAD, circuit design, PLC programming and industrial electrical installations.
Certifications and Schools
CSC Deep Learning and GPU Programming Workshop 2020
- NVIDIA Deep Learning Institute: Accelerated Computing with CUDA C/C++
- NVIDIA Deep Learning Institute: Deep Learning for Multi-GPUs
CSC Summer School in High Performance Computing 2019
- HPC programming with message-passing, threading and parallel I/O using C, Fortran, OpenMP, Open MPI and OpenACC.
Research
Peer Reviewed Journal Publications
- Bioplausible synaptic behavior of Al/GdCaMnO/Au memristive devices for unsupervised spiking neural networks
- Compact modeling and SPICE simulation of GCMO-based resistive switching devices
- Electron doping effect in resistive switching properties of Al/GdCaMnO/Au memristor devices
- Tuned AFM-FM coupling by the formation of vacancy complex in GdCaMnO thin film lattice
- Metastable ferromagnetic flux closure-type domains in strain relaxed GdCaMnO thin films
- Appearance of glassy ferromagnetic behavior in GdCaMnO(0<<1) thin films
- Transport properties of resistive switching in Ag/PrCaMnO/Al thin film structures
News
- Linear, Significant funding for Linear’s AI research, 2025
- Yle, The patented memristor is so small it fits even on a phone’s microchip, 2024 (in Finnish)
- Uusi Teknologia, Memristor won the engineers’ invention competition, 2024 (in Finnish)
- Tekniikka & Talous, In Finnish research, computers imitate the human brain: memory state survives even a power outage, 2022 (in Finnish)
- Talouselämä, Finnish company created an AI that writes property listings on behalf of real estate agents, 2022 (in Finnish)
- Turun Sanomat, University of Turku receives 635,000 euro Business Finland funding for AI research, 2022 (in Finnish)
- Aamuset, Physics Days demonstrated how nitrogen boils, 2018 (in Finnish)
Patents
- Novel thin film material for memristor, and a memristor comprising such material, P. Paturi, A. Schulman, H. Huhtinen, V. Lähteenlahti, A. Beiranvand, FI20205101 · PCT/FI2021/050057 · TW110103063 · US 17/796,819, 2020
Conferences
- International Conference on Memristive Materials, Devices & Systems (MEMRISYS), Poster presentation, Dresden, 2019
- 12th Spanish Conference on Electron Devices, Poster presentation, Salamanca, 2018
- Physics Days 2018, Poster presentation, Turku, 2018
Misc.
- Positron annihilation spectroscopy beamline experimenter, ELBE; Helmholtz-Zentrum Dresden Rossendorf, Dresden, 2020
- Presentation, Unsupervised learning of digit recognition using spike-timing-dependent plasticity, Machine learning and algorithmics seminar, University of Turku, 2019
- Video, 3D reconstruction from old photos using neural radiance fields, 2022
Skills
- Machine learning: XGBoost, PyTorch, scikit-learn; LLM systems (agents, RAG, structured outputs), vision and autoencoder models, geospatial prediction services; data/model evaluation and validation.
- Engineering: Python, PostgreSQL, Azure, AWS, Linux, REST APIs, containers, CI/CD, DevOps, testing, architecture; full-stack development; agile methods.
-
Languages
- Python (PyTorch, scikit-learn, FastAPI)
- SQL
- C
Scientific
- Physics: simulations, modeling, experiment design, device instrumentation; instrument control and data-acquisition software.
- Materials and devices: resistive switching, perovskite oxides, neuromorphic devices; thin-film nanofabrication.
- Methods: PLD, XRD, EBPVD, SEM, AFM, XPS, SQUID, PPMS; low-temperature measurement and cryogenics.





