Case Study: Building a Machine Learning Engineering Team for a European Insurtech
TechNest Talent | AI and Machine Learning | Contract and Permanent Recruitment
The Client
A Series B insurtech headquartered in Munich with 180 employees, operating across Germany, the Netherlands, and the UK. The company uses machine learning models to automate claims processing, detect fraud in real time, and personalise policy pricing for commercial insurance products.
The Challenge
The client's existing data science team of four had built proof-of-concept models for claims automation and fraud detection, but none had reached production. The models worked in notebooks. They did not work at scale.
The VP of Engineering identified three problems:
- No MLOps capability. There was no model deployment pipeline, no monitoring, no automated retraining, and no feature store. Every model was a manual, one-off exercise.
- A senior leadership gap. The data science team reported into a software engineering manager who had no machine learning experience. Technical decisions were being made without the right expertise.
- A hiring timeline that did not match the business plan. The board had approved a twelve-month roadmap to take all three ML products to production. The team needed to grow from four to fourteen within that period, across three countries, mixing permanent and contract hires.
The client had spent four months working with two generalist recruitment agencies. Neither had produced a single hire. The candidates they submitted were either data analysts mislabelled as machine learning engineers or senior professionals who withdrew after discovering the recruiter had misrepresented the role.
What TechNest Talent Delivered
Phase 1: Contract MLOps Engineers (Weeks 1 to 4)
The most urgent need was MLOps. Without deployment infrastructure, nothing else on the roadmap could progress.
We placed two contract MLOps engineers within nine working days of the initial briefing:
- Contractor 1 — Based in Berlin. Five years of MLOps experience across AWS SageMaker and Kubeflow. Previously built the ML platform for a Series C fintech in Frankfurt. Outside-scope ANuG assessment completed before contract signature. Started on day twelve.
- Contractor 2 — Based in Amsterdam, working remotely. Specialist in ML monitoring and automated retraining pipelines using MLflow and Weights and Biases. DBA compliance handled by TechNest Talent. Started on day fourteen.
Within eight weeks, the two contractors had built a functioning ML deployment pipeline, implemented model versioning, and deployed the claims automation model into a staging environment for the first time.
Phase 2: Permanent Head of Machine Learning (Weeks 2 to 10)
Alongside the contract placements, we ran a confidential headhunt for a permanent Head of Machine Learning to lead the function long term.
This was not a role we could fill from job board applicants. The client needed someone who had scaled an ML team from single digits to double digits, had experience in regulated financial services, spoke fluent German and English, and was willing to be based in Munich with flexibility for remote work.
We mapped 34 potential candidates across Germany, Switzerland, and Austria. We approached 19 directly. Seven entered the process. The client interviewed four and made an offer to the strongest candidate, a former ML Director at a mid-cap insurance group in Zurich who was looking to move back to Germany for family reasons.
Time from briefing to accepted offer: 8 weeks. Notice period managed: 3 months (standard for Swiss senior hires). TechNest Talent coordinated the transition timeline so the contract MLOps engineers provided continuity until the new Head of ML started.
Phase 3: Scaling the Permanent Team (Months 3 to 11)
Once the Head of ML was in place, we worked with them directly to build out the permanent team across three locations:
| Role | Location | Time to hire |
|---|---|---|
| Senior Machine Learning Engineer | Munich | 4 weeks |
| Machine Learning Engineer | Munich | 3 weeks |
| NLP Engineer | London (remote) | 5 weeks |
| Data Scientist (Fraud) | Amsterdam | 6 weeks |
| Data Scientist (Pricing) | Munich | 4 weeks |
| Computer Vision Engineer | Berlin (remote) | 7 weeks |
| ML Platform Engineer | Amsterdam | 5 weeks |
| Junior Data Scientist | Munich | 3 weeks |
Eight permanent hires across three countries in nine months. Every hire was still in post twelve months after joining.
Phase 4: Contract Extension and Transition (Month 10 onwards)
Contractor 1 was extended for a further six months to lead the fraud detection model deployment, which had become more complex than originally scoped. Contractor 2 completed their engagement on schedule and transitioned their work to the newly hired permanent ML Platform Engineer, with a two-week handover period managed by TechNest Talent.
The Outcome
Before TechNest Talent:
- 4-person data science team
- Zero models in production
- No MLOps infrastructure
- No ML leadership
- 4 months spent with generalist agencies, zero hires
After twelve months with TechNest Talent:
- 14-person ML function across three countries
- 1 permanent Head of Machine Learning
- 8 permanent hires, all retained at twelve months
- 2 contract MLOps specialists
- Claims automation model live in production (processing 12,000 claims per month)
- Fraud detection model in final testing
- Pricing personalisation model in staging
- Full ML deployment pipeline with automated monitoring and retraining
Client Testimonial
"We wasted four months with agencies that did not understand what we were building. TechNest Talent placed two contractors in under two weeks who built the infrastructure our internal team had been unable to deliver in six months. Then they found us an ML Head we would never have reached through a job advert. A year later, we have a fourteen-person team, three models either live or near-live, and zero regrets about the hiring decisions. They understood the technical landscape, the urgency, and the complexity of Europe. We did not have to explain what MLOps was."
— VP of Engineering
