Regional AI Lead · HiQ · Västerås
I lead HiQ's regional AI practice in Västerås, defining AI strategy and governance, delivering Generative AI and LLM solutions on Azure and Databricks, and building the teams that ship them. My background is in autonomous vehicles, perception, and safety-critical systems, so I care as much about the engineering as the strategy.
Where AI meets industrial engineering.
I lead HiQ's regional AI practice in Västerås, where I build and scale AI operations for enterprise clients. My work spans defining AI strategy and governance frameworks, delivering Generative AI and LLM solutions on Azure and Databricks, and recruiting and developing AI talent.
Current projects include designing cloud-based data and AI platforms with lakehouse architectures, and applying LLMs to analyze maintenance data for industrial clients. I focus on turning real engineering and operational challenges into practical, deployed AI solutions, not prototypes that stall.
Before HiQ, I spent years as a system architect and engineer in industrial and automotive settings (Scania, Volvo Autonomous Solutions, Veoneer), working on data pipelines, sensor fusion, and safety-critical systems. Earlier, I built AI capabilities from scratch in startups, covering everything from neural networks and computer vision to embedded systems and FPGA design.
MSc in Robotics & Mechatronics, Mälardalen University. Studies in Computer Vision & Mathematics, Uppsala University.
From FPGAs and robotics to autonomous vehicles and enterprise AI.
Build and scale HiQ's regional AI practice: AI strategy and governance, Generative AI & LLM delivery on Azure and Databricks, cloud data/AI platforms with lakehouse architectures, and growing the local team.
Designed and integrated autonomous vehicle technologies for hub-to-hub logistics; architected scalable data pipelines and real-time analytics for large sensor-data streams, balancing rapid MVPs with long-term scale.
Signal processing and sensor fusion for autonomous vehicles: object detection, localization, and free-space detection using radar, camera, and LIDAR data, validated in simulation and real vehicle testing.
Developed AI applications for computer-vision-based safety systems (ADAS / automated driving) in the automotive industry.
Led development of AI-driven signal analysis for hand-tracking input devices (detection, tracking, and estimation), plus FPGA/VHDL work and C/C++ backend and frontend interfaces.
Led the control system for FUMO, a firefighting ROV, and built AI-driven computer vision for autonomous decision-making and situational awareness; later applied computer vision to precision agriculture.
How I help teams and clients.
Define where AI creates real value, and the governance frameworks that keep it safe, compliant, and maintainable as it scales across an organization.
Design and deliver LLM solutions on Azure and Databricks, from analyzing industrial maintenance data to production-grade assistants and pipelines.
Cloud data and AI platforms built on lakehouse architectures: the foundation that turns raw operational data into deployable models.
Deep experience in autonomous systems: object detection, localization, and sensor fusion across radar, camera, and LIDAR for safety-critical use.
Off the clock: embedded firmware and home energy at home.
A whole-home energy optimizer coordinating a solar-plus-battery system, EV charging, and pool equipment against Nord Pool 15-minute prices to minimize electricity cost automatically.
ESP32 firmware controlling both the pool pump (continuous RPM) and the heat pump over RS485/Modbus, plus a planning layer on top. A machine-learning model learns the pool's thermal behaviour over time, estimating heat dispersion and losses, to plan when to circulate and when to heat. You tell it when you'll use the pool; it schedules water at the right temperature for then, at the lowest cost, and manages water chemistry, suggesting when to dose chlorine, adjust pH, and clean the filter.
Device-free presence detection and tracking from WiFi Channel State Information, using the multipath changes a human body causes in ordinary WiFi signals. An ESP32-S3 extracts per-subcarrier CSI and ML models turn it into presence, motion, and people counting, with no cameras, no wearables, works in the dark and through walls.
Ten things I worked out about AI this summer.
Agents turn AI into something you delegate to, not something you query; the bottleneck is no longer what the model can do but what you dare to hand over, and governance is what makes the boldness possible.
Doing nothing is not the absence of risk but a deferred one nobody books; on legacy, coding agents, and why standing still is usually the more expensive choice.
The more capable the system, the less you can prove it's right; the answer isn't chasing guarantees but measuring continuously and building containment around what can't be verified.
Cultivating experimentation and controlling cost aren't opposites but three different guardrails: competence that lifts with learning, an absolute cap that protects the whole, and an architecture that makes every action cheap.
The only honest defense is one that's constantly attacked from within; but whoever cultivates offensive capability has built a weapon in the house and must hold it on a tighter leash than anything else.
Clear rules beat unclear ones no matter how strict, and whoever built by the rules paid up front; unregulated speed takes you fast to the places that weren't worth reaching.
Consolidation into one end-to-end model wins where module boundaries destroyed more information than they protected, but it works only where the safety envelope is simpler than the function itself.
Whoever owns the cheap, clean, abundant power ends up owning the compute, the labs and the models; the West is about to repeat the exact dependency mistake it made with Russian gas and fertilizer.
Physical AI's breakthrough isn't better bodies but a brain that finally handles the unpredictable real world; the point isn't compassion for the weak but enabling, returning autonomy to people.
The human role in the age of AI isn't to be smarter than the machine but to be the one who carries responsibility for what it does; we're no longer the best decisions, only the ones who can be held accountable.