Hard problems, shipped
Each case follows the same pattern: the problem the client had, what we built, and what changed. Results are what Nordic buyers look for first.
Four recent cases
ML/LLM management platform
Problem: teams could train models but not run them reliably.
What we built: a SaaS platform covering the whole model lifecycle: data preparation, training, tuning, versioning, CI/CD, monitoring and team collaboration.
Stack: Python, R, Java, React, Kubernetes, AWS, GCP, TensorFlow, PyTorch, Spark, Kubeflow, Metaflow, Vertex AI.
[Result: users, models in production, time saved? Client work or own product? Client name if allowed.]
Food product modelling platform
Problem: food labs spent months testing new recipes physically.
What we built: the digital side of the lab. ML and LLM tools that model flavour, texture, nutrition and cost before a prototype is made.
Stack: Python, React, GCP, Keras, XGBoost, PyTorch, Spark.
[Result: fewer physical trials? faster time to market? Client sector and country.]
Moving-object identification
Problem: identifying objects from radio-frequency sensor data in real time.
What we built: a sensor-fusion platform that classifies objects from RF spectrum data and predicts their movement.
Stack: Python, Java, React, Kubernetes, Kafka, TensorFlow, Keras, XGBoost.
[What can be said publicly about client and use? Accuracy or latency achieved?]
Quantum-aware AI (QaaS)
What we built: hybrid AI and quantum workflows for optimisation, risk simulation and medical imaging, with LLM, RAG and agentic layers.
Stack: IBM Quantum, Qiskit, AWS, Python, React.
[Is there a client? What problem did it solve better than classical methods?]
Earlier milestones
- 2003: first major project for DaimlerChrysler Bank.
- 2008: 50 native mobile apps launched.
- e-Government CRM for Swiss authorities.
- Inspection apps for Dutch seaports.
- Situational awareness system for transit authorities.
- Healthcare inventory and compliance system.
[Which of these may be named publicly?]
How a Nordic engagement could look
Invented stories to show the format. NewGround: replace each one with a real project told the same way.
EXAMPLE
Predictive maintenance for a Danish industrial manufacturer
Sensor data from 40,000 installed units sat unused. A team of four built a data pipeline and an anomaly model in twelve weeks. Service visits are now planned from the data, not the calendar.
EXAMPLE
Legacy migration for a Nordic pension provider
A 20-year-old Java core with three people left who understood it. We documented, wrapped and migrated it module by module over 18 months, with zero unplanned downtime.
EXAMPLE
From RAG pilot to production at a Swedish health-tech scale-up
An internal knowledge assistant worked for ten testers but broke at 500 users. We rebuilt retrieval, added evaluation and monitoring, and handed over a platform the in-house team now runs.
