Andreas Ulvig
AI / Machine Learning Developer
ML & computer vision · NLP · prototypes & research
"Sitater er ikke mitt fagfelt"
At a glance: Python · Docker · BERT · reinforcement learning · deep learning · Label Studio · Cursor · GitLab
AI civil engineer (UiA, graduated Jan 2026) with hands-on experience in ML, computer vision, and NLP. Co-founder and developer at Acolight; internships at VOCA, Norkart, and Xtreme Controls; teaching and research assistant roles at UiA. Strong in Python, MLOps, and AI-assisted development (Cursor); experience across prototypes and research projects.
Key qualifications
- Enthusiastic, analytical, solution-oriented. Primary language: Python. C++ from internship and projects; familiar from coursework and projects: JavaScript, Kotlin, Scala, Dart, R
- Bachelor thesis: synthetic data for object detection (Unreal Engine, Python, YOLOv5). Master thesis: autonomous drone navigation with Soft Actor-Critic (SAC) in Unreal Engine and AirSim
- 3 years on the board of Beta; central role in rebuilding the corporate committee post-pandemic; ~70 credits of extra courses
Education
University of Agder (UiA), Grimstad, Norway
- Pre-course for engineering students (Aug 2018 – Jun 2019)
- Master thesis: autonomous drone navigation with Soft Actor-Critic (SAC) in Unreal Engine and AirSim
- Graduated January 2026
Skills
- DevOps / MLOps
- GitLab, Docker, Label Studio
- AI-assisted development
- Cursor, AI coding assistants
- Databases
- SQL, Neo4j, MongoDB, Firebase (student projects)
- Programming
- Python (primary). C++ (internship & projects). JavaScript, Kotlin, Scala, Dart, R (coursework & projects)
- Languages
- Norwegian, English (fluent written & spoken)
Work experience
Acolight, Kristiansand, Norway
- Delivered paid RAG-based LLM prototype to client for knowledge retrieval. Now developing an in-house prototype (Acolight-owned); AI elements planned in the long run
- Prototypes, implementation plans, and data overviews to pitch ideas to specific potential customers
University of Agder (UiA), Grimstad, Norway
- Teaching assistant: Application Development (twice), Applied Algorithms, Introduction to AI
- Research: translating and supporting a survey on technology use in healthcare (collaboration with nursing students); UiA CLEAR project
Norkart, Kristiansand, Norway
- Research: implemented a generative BERT variant trained with SAC reinforcement learning; documented learning via statistical analysis
- Supported other students e.g. in object recognition tools for document classification
Xtreme Controls, Basingstoke, UK (Remote)
- ML-based prototype for hotel room temperature estimation (blinds open/closed) for energy saving; Dockerised and tested
- Two neural nets predicting future temperature; structured data from MongoDB and external weather APIs
VOCA, Kristiansand, Norway
- Autonomous cargo handling: steps towards self-driving forklift; achieved simple autonomous driving by end of period
- Communication between camera and mapping algorithm (GStreamer); hands-on experience with CMake
Kristiansand, Norway
- Details on request if relevant
Other roles: Member → leader of bedkom → deputy chair of Beta; board member, deputy chair and chair of Sorlandet-BD; coordinator of fif (meeting organisation, communication between student organisations).
Selected projects
- Deep reinforcement learning for autonomous drone navigation in Unreal Engine and AirSim using SAC; high-fidelity simulation for safe, scalable training with relevance to real-world autonomy and robotics
- Active learning substantially reduced manual labelling effort (~100 labelled examples for good classifier performance); BERT-based classifier integrated with Label Studio. Pipeline ingests data, supports labelling, and surfaces sentiment on an interactive map for stakeholders
- Cross-disciplinary project with Kristiansand harbour: RL prototype for berth assignment; Plotly visualisation; sole technical contributor
- Ensemble of U-Nets; implementation aligned with MapAI competition style; two-person group
- Synthetic data for object detection (Unreal, YOLO); LSTM stock prediction (StockNet); PPO vs SAC on Bipedal Walker; Tsetlin Machine for hate-speech detection; contactless heart-rate monitoring; generative BERT with SAC; SQL vs Neo4j vs MongoDB for data warehousing; hand gestures (classification); Kotlin/Android and Flutter/Dart apps.
- Code on GitLab; most repos private — access on request.