Hi, I'm Jacob.
I build ML-powered software.
Machine learning engineer working across computer vision, applied machine learning and full-stack cloud platforms — from training segmentation models to shipping production apps on Kubernetes-grade cloud infrastructure.
Jacob Salomonsen
Machine Learning Engineer
About
Engineer, builder, lifelong learner.
I'm an AI & Machine Learning engineer with over a decade of experience across the full ML lifecycle — from LLMs and computer vision to time-series forecasting and large-scale behavioral analytics, built on Python, PyTorch and Spark within AWS and Kubernetes environments.
Professionally I've shipped automated ML platforms at Apple and E.ON, run behavioral analytics on Spark at Sony Mobile, and built data pipelines at Twilio and Steria — work that spans everything from SVMs to transformers, and from data annotation through to production deployment and continuous model updates. As founder of Nordlytics AB I now deliver end-to-end AI solutions for Fortune 500 and utility clients.
On the side I build full-stack ML products like Shot Log — a computer-vision target-scoring app (instance segmentation, model distillation and ONNX inference on the edge) deployed to DigitalOcean with Terraform and GitHub Actions — and Indoor Environment AI, an IoT closed loop where ESP32 sensors feed InfluxDB and an hourly scikit-learn scheduler with SHAP explanations controls a dehumidifier.
Years building ML systems
Enterprise clients served
Responsibility & ownership
Autonomous production systems
Projects
Things I'm building.
A selection of personal projects. Each card links to a deeper write-up with architecture and the technologies behind it.
Shot Log & Shot Detection
Computer-vision instance segmentation (homography + impact detection) for automatic target scoring, chained into a FastAPI + React Native shooting-log app deployed on DigitalOcean.
Indoor Environment AI
Pimoroni Enviro + Shelly sensors stream into InfluxDB via Telegraf; a FastAPI backend runs an hourly scheduler that pulls readings, runs a scikit-learn humidity regressor (with SHAP explanations) and threshold- decides whether to switch a dehumidifier on or off — all managed from a React dashboard.
Shot Gyro
Web HID app streams DualSense gyroscope and accelerometer data at ~125 Hz into CSV, Label Studio annotates shooting phases (Rest, Lift, Aiming, Shot, Follow), and a ruptures change-point pipeline segments each shot from the raw motion signal.
Product Configurator
Java EE 6 (EJB 3.1 + JAX-RS/RESTEasy + JPA/Hibernate) backend deployed as an EAR to JBoss AS 7, backed by MySQL (configuration graph) and MongoDB GridFS (3D model assets), serving a Backbone + Three.js editor and an embeddable configurator widget.
Career
Achievements & milestones.
Highlights from my professional journey so far.
Full-Stack AI Webapp & Edge Inference
Leading end-to-end development of a computer vision webapp for a non-disclosed client through Nordlytics, owning infrastructure (Terraform on AWS), MLflow-managed model lifecycle, and edge inference on thin devices.
Founder & CEO, Nordlytics AB
Founded an independent AI & Machine Learning consultancy delivering high-impact automated solutions for Fortune 500 and utility clients across the full ML lifecycle.
Senior ML Engineer, Apple
Built fully automated ML/AI pipelines that clean, combine and augment data from many diverse sources, replacing a labor-intensive manual process with remarkable, measurable improvements — from conception and annotation through production deployment and continuous model updates.
Automated Forecasting Platform, E.ON
Owned creation of a fully automated time series forecasting solution for critical power-grid points, spanning data ingest, feature generation, and training/running of models from OLS to Seq-to-Seq quantile regression, with metrics and results surfaced in a UI.
Large-Scale Behavioral Analytics, Sony Mobile
Performed comprehensive user behavioral analysis on large datasets using Apache Spark (PySpark) — data cleaning, clustering, time series forecasting, and classical/deep-learning models powering regular behavioral model updates.
Lead Developer, Steria
Development and maintenance of the import/export taxation system for the Danish tax ministry, acting as lead developer with shared project and team management responsibilities.
MSc Computer Science, Copenhagen University
Coursework in Machine Learning, Statistics, and Image Analysis; research implementing the lattice Boltzmann method using CUDA and GPUs.