lvl100Sign in

prosus

Principal Applied AI Scientist, Commerce Foundation Models

  • Amsterdam
  • Full-time
  • Posted

Skills

  • Communication
  • Large Language Models
  • Leadership
  • Machine Learning
  • Python
  • Statistics

Job description

Who We Are

You've ordered dinner through Just Eat Takeaway. You've bought and sold on OLX. You've paid with PayU. Chances are, you've already used a Prosus company today.

We are Prosus, a global technology company powering some of the world's leading lifestyle ecommerce businesses. Every day, billions of people across India, Latin America, Europe and beyond use our products and services to buy, sell, pay, connect and discover.

Our AI-driven ecosystem spans 100+ companies across ecommerce, food delivery, fintech and experiences, serving billions of consumers worldwide. AI powers how we operate and innovate, with more than 60,000 active AI agents helping our teams build, scale and move faster. Through our investment team, we also back the founders and technologies shaping the future: From robotics and autonomous systems to synthetic biology and frontier AI.

Guided by our five values Innovation, Results, People, Impact, and Entrepreneurship, culture is our competitive advantage. We thrive with experimentation; we build, we test, we learn, we iterate.

Join us in shaping the future of technology at global scale.

ABOUT THE TEAM

Join the AI team at Prosus, a global technology group with businesses in food delivery, ecommerce, classifieds, travel and payments. We build LCM, the Large Commerce Model: our own foundation model, trained on billions of transactions and interactions from proprietary data across the portfolio. LCM already powers search, recommendations and marketing in production at iFood. We are extending it across our companies around the world and to businesses outside the group, while broadening its capabilities in agents and real-time prediction tasks. Your work will shape what LCM learns next and feed directly into products used by millions of people.

ABOUT THE ROLE

We’re looking for an experienced applied scientist to help set LCM’s research direction and work directly on its most demanding scientific problems. This is a hands-on individual contributor role in a flat team of around 30 engineers, with no direct reports. You will mentor colleagues and contribute to hiring, but you will not manage people.

We define how a capability will be evaluated before developing it, and use that evidence to guide training and deployment decisions. You will help decide which research questions deserve investment, design and run experiments, and develop training methods that can be used across applications.

The role requires deep expertise in a core area of model training and the judgment to connect that work with decisions about data, evaluation and deployment. Scientists and engineers work together from the first experiment through to production evaluation, and you will take scientific responsibility for agreed research projects.

A central research goal is to learn from a continuous stream of billions of user sessions across the portfolio and turn useful signals into regular updates to model weights.

WHAT THE ROLE INVOLVES

•Develop the research agenda with the team, identifying questions that affect model quality, training efficiency or downstream results. Use experiments to decide which approaches warrant larger training runs and which should be revised or stopped.
•Lead selected research projects through implementation, training and evaluation, staying directly involved in writing code, inspecting data, debugging runs and analysing failures.
•Establish evaluation methods that separate real improvements from leakage, noise and benchmark overfitting, and work with portfolio teams to test whether offline gains hold in controlled online experiments.
•Mentor scientists and engineers through joint research, experiment reviews and technical discussion, and advise on work that crosses project boundaries.
•Contribute to hiring, publish research and support the release of models, benchmarks and training methods where the work and data can be shared.

RESEARCH AREAS

The initial focus will depend on your expertise and the team’s priorities. We expect substantial depth in one of the two areas below, with the ability to contribute to decisions across the other.

Pre-training and representations

We continue pre-training open-weight models on commerce data. Research questions include how to represent users and catalogues, how to combine data from different businesses, and how to allocate compute effectively. Work may include:

•Studying data mixtures, scaling behaviour and training schedules to guide the next run.
•Designing sequences from user histories, merchant catalogues and market data, including alignment between text and semantic IDs for users, items and merchants.
•Developing mid-training methods for capability acquisition, context extension and retention of general capabilities.
•Evaluating synthetic data and distillation methods through controlled experiments, including the trade-offs between quality, inference cost and latency.
•Studying transfer across languages, markets and commerce categories, with evaluations for new users and items and for performance on future data.
•Tabular foundation models are a possible specialisation, including pre-training and in-context learning on structured commerce data.

Post-training and continual learning

We train LCM for commerce tasks and a range of agent applications, including customer-facing assistants and agents that carry out business operations. We also develop continual learning methods that turn new interactions into regular updates to model weights. Work may include:

•Developing supervised fine-tuning, preference optimisation and reinforcement learning methods, using verifiable outcomes or learned reward models.
•Turning session histories, searches, clicks, purchases, repeat orders and operator feedback into training examples and rewards, accounting for exposure bias, delayed feedback and the difference between observed engagement and incremental business value.
•Studying data selection, replay and update frequency so models learn from new interactions without losing established capabilities.
•Training and evaluating decision models that choose rankings, offers and promotions, using offline learning and off-policy evaluation to estimate policy value and uncertainty and inform controlled online experiments.
•Training models for multi-turn conversations and multi-step agent workflows, including understanding user intent, planning, tool use, maintaining context, recovering from errors and verifying outcomes. This includes building RL environments with simulated users and catalogues, and testing whether gains transfer to real interactions within defined permissions and business rules.
•Calibrating reward models and LLM-based graders against expert judgments and observed outcomes, and investigating reward hacking and other unintended effects.
•Developing automated experimentation methods that propose and test changes to training recipes, prompts, tools or agent control flow, with held-out evaluations and regression checks.

HOW WE EVALUATE THE WORK

Research should produce evidence that supports a training or deployment decision. We use strong baselines, including existing production systems, and weigh model quality against compute requirements, inference cost and latency.

Experiments should be reproducible, with ablations, uncertainty estimates and failure analysis. They must account for temporal leakage, changes in the data distribution and differences between companies and markets. Negative results are valuable when they explain a limitation or stop further investment in an approach that does not work.

The work also involves data provenance, permitted use across companies and privacy-aware handling of customer histories, together with our data and engineering teams.

In the first three months, we expect you to establish a focused research programme, resolve an important training or evaluation question, and work with the relevant teams to test the resulting method in a downstream application.

MINIMUM QUALIFICATIONS

•A PhD or equivalent industry research experience in machine learning, computer science, statistics or a related field.
•At least eight years of research or applied science experience in machine learning, including doctoral research, with at least four years leading research directions that were published or reached production.
•Substantial experience developing and training models, with deep expertise in pre-training, post-training or reinforcement learning relevant to the areas above.
•A record of original contributions, shown through research papers, released models, benchmarks, open-source projects or significant advances in deployed systems.
•Scientific leadership: choosing research directions, resolving difficult technical questions and helping experienced colleagues make better decisions.
•Sound experimental and statistical judgment, including examples where the evidence changed your approach and the ability to recognise when the evidence does not support a conclusion.
•Strong Python and PyTorch skills and the ability to work independently in a shared research codebase. Research on large language models requires hands-on experience with distributed training.
•Clear communication and a collaborative approach to disagreement, review and mentoring.

PREFERRED QUALIFICATIONS

•Experience with recommender systems, generative retrieval, semantic IDs, continual learning, synthetic data, distillation or tabular foundation models.
•Familiarity with commerce data and online experimentation.
•We do not expect expertise in every area.

WHAT WE OFFER

•Influence over LCM’s research agenda and which questions the team invests in.
•Proprietary commerce data, B300 and H200 GPUs for large-scale training, and external models for evaluation and benchmarking.
•The opportunity to take research from early experiments through publication and deployment, working directly with portfolio teams on products used by millions of people.
•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.
•An uncapped budget to recruit and mentor research interns for your projects, with access to candidates through our connections with leading universities.
•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.

HOW TO APPLY

Please include examples of research or deployed work that show your contribution, the decisions involved and the evidence used to evaluate the results, and indicate which research area best matches your depth.

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.

Browse all jobs by role, city or company