ABOUT THE ROLE
Evaluation is the core of everything we do. We build from the evals: we define how success will be measured before developing a capability, and use the results to guide training, model selection and deployment. We already have hundreds of evaluations spanning domain knowledge, commerce and gaming, with plans to expand their coverage, depth and scale.
This work builds on ProLLM, the public benchmarking platform we ran for real-world use cases until our coding assistant benchmarks became saturated. Since then it has been a private benchmarking platform for Prosus companies. We plan to open it again to publish our commerce benchmarks, covering both agentic and decision-making tasks.
We’re looking for a Senior Machine Learning Engineer to own the development of these evaluations and the platform that runs them. The work involves defining meaningful tasks, curating evaluation data, implementing reliable scoring, investigating model behaviour and making results useful for research and product decisions.
In our flat team of around 30 engineers, you’ll work directly with applied scientists, training engineers and domain experts. You’ll help teams contribute evaluations and build the infrastructure and automation needed to run them throughout model development.
In the first three months, we expect you to take technical ownership of part of the evaluation platform, deliver a set of new or improved evaluations with curated data and calibrated difficulty, and work with scientists to use them in a training or release decision.
WHAT THE ROLE INVOLVES
Design evaluations with scientists and domain experts, translating intended capabilities into representative tasks, datasets and success criteria. Cover current applications and capabilities we want future models to acquire.
Curate evaluation data and calibrate task difficulty, so that evaluations are neither saturated nor out of reach and still separate models of different quality. Work with domain experts on sourcing, annotation and review, within data provenance, permitted use across companies and privacy requirements.
Own the development and operation of the evaluation platform across training checkpoints, deployed models and external baselines. Support distributed execution, reproducible configurations, trace inspection and comparison of results, while managing runtime and cost.
Develop programmatic checks, outcome-based scoring and LLM-based graders. Calibrate automated judgments against expert reviews, investigate disagreements and test whether graders reward superficial answers or reject valid solutions.
Analyse model outputs and agent trajectories, including tool calls, environment state and task outcomes. Distinguish model failures from problems in the evaluation, quantify uncertainty and explain what changed between versions.
Maintain versioned tasks, datasets, graders and environments, with protected held-out tests and contamination checks. Expand coverage as models improve, retaining useful regression tests and adding harder tasks where existing tests no longer distinguish performance.
Prepare commerce benchmarks for public release on ProLLM, including public mirror sets and evaluation code, where the data can be shared.
Integrate evaluations into training and release workflows, with dashboards and regression checks that support concrete decisions. Work with portfolio teams to establish whether offline improvements correspond to better outcomes in online experiments.
Build tools and agents that help generate candidate tasks, run evaluations and investigate failures, validating their outputs before adoption. Review contributions and mentor engineers so teams can extend the suite consistently.
MINIMUM QUALIFICATIONS
At least five years of ML or software engineering experience, including substantial hands-on work building and evaluating ML systems.
Experience owning evaluation frameworks, benchmarks or model-quality systems that informed training or deployment decisions.
Strong Python and experience with PyTorch, Transformers or comparable ML libraries, alongside reliable production software development.
Practical knowledge of experimental design and statistics, including sampling, uncertainty, repeated trials and comparisons across model versions.
Experience running inference or evaluation workloads at scale and debugging failures across models, data, scoring code and infrastructure.
Clear communication and the ability to turn ambiguous results into defensible recommendations.
PREFERRED QUALIFICATIONS
Experience with language-model training or post-training.
Experience with agent or gaming environments.
Experience with human evaluation, LLM-as-judge calibration, recommendation metrics or online experimentation.
Contributions to evaluation frameworks, public benchmarks or research.
WHAT WE OFFER
Ownership of the evaluation platform and the evaluations that guide LCM’s research, training and deployment.
Access to proprietary domain data, GPU infrastructure and external models for comparison.
Direct collaboration with researchers, engineers and portfolio teams, with scope to publish methods and benchmarks, including through ProLLM, where the work can be shared.
Colleagues who have released Hugging Face models with more than 100 million downloads, authored papers at NeurIPS, created benchmarks and built multiple production AI systems.
Current ML frameworks and unlimited access to coding assistants.
A base at our Amsterdam office and AI House, with relocation support. We also consider remote arrangements for candidates with particularly strong experience relevant to the role.
Competitive compensation and a top-spec MacBook Pro.
If you want to build the evaluations that shape what a foundation model learns, we’d like to hear from you. Please include examples of evaluations or platforms you have built, how you established that their results were reliable, and the decisions those results informed.
Why Prosus?
At Prosus, your colleagues include 60,000 AI agents… and your job is to figure out what that actually means. We're one of the largest operators of agentic AI on the planet, inside a $100B+ portfolio across 100+ companies and two billion people. The humans here own what AI can't do alone: judgment, creativity, and the call on what gets built next.
Own the impact. Work the ecosystem. Earn your growth. Build with backing.
Curious about our perks and benefits? Take a look at this page https://prosusbenefits.com/.
Our Diversity & Inclusion Commitment
We respect the dignity and human rights of individuals and communities wherever we operate in the world. Building an inclusive workplace where everyone feels welcome and can thrive is critical for us. We provide access to education, which helps everyone understand the important role they play and the positive impact they can have.
Want to know more about life at Prosus? Explore our culture and values https://www.prosus.com/the-prosus-way.