Sourcein

AI/ML Engineer - Synthetic Data Platform

 

The problem is genuinely hard

How do you model how real people answer questions? Not fake data, not sampled noise - a statistically valid simulation of human survey behaviour across hundreds of millions of respondents in 150 countries.

That's what we're building. And we haven't solved it yet.

This is a greenfield project at the intersection of classical ML, statistical modelling, and frontier LLM techniques. Customer pilots are already running. The roadmap is shaped by real feedback. The best approach will emerge from the research - your research.

 

What the work looks like

* Design and run experiments for survey augmentation and audience profiling models
* Fine-tune open-weights LLMs (LoRA, distillation) or build custom architectures - whatever gets the best result
* Own the full path from experiment to production
* Partner with a dedicated data science team on statistical rigour and experiment design
* Use AI-assisted development tools to move fast

 

What we're looking for

* Strong classical ML and statistical foundations - probability, statistics, the fundamentals
* Real LLM experience - fine-tuning, evaluation pipelines, model architecture. Not API wrappers
* Production instincts - you ship what you build
* Comfortable with ambiguity - the best solution isn't obvious yet

 

Academic or research backgrounds welcome. If you'd rather ship than publish, this might be the right move.

 

What makes this intesting

* Genuinely unsolved problem at industry scale
* Full technology flexibility - use whatever works
* Small team, high autonomy, real customer impact from day one
* Fully remote within Czechia - Prague office available if you want it

Data

Research

Data Science

Statistics

Roadmap planning

MLflow

Statistical model

Lora

Experimental Design

Probability

AI / ML / LLM Research Engineer