Member of Technical Staff (Business Foundation Models)

About Pavo

Pavo is building the Applied Science Factory: a system that develops an understanding of production systems, investigates performance, constructs reliable evaluations, and helps engineers improve business outcomes.

We work with the underlying product, code, data, models, and experiment history. Our research spans training environments, scientific agent capabilities, and models for structured and behavioral data.

We're a small team with experience building large-scale ML systems at Spotify, ShareChat, and Sourcegraph. Members of Technical Staff own research and implementation end to end.

The role

Pavo works on production problems involving user journeys, transactions, subscriptions, consumption, model decisions, and business metrics. Many of these tasks use related signals but differ in target definitions, schemas, and available history.

You'll investigate models that reuse learning across datasets and prediction tasks. The work spans tabular and relational data, temporal event sequences, pretraining, adaptation, and calibrated prediction. Example applications include predicting future user behavior, retention, purchase patterns, and other business outcomes.

You'll own research design, data preparation, model training, evaluation, and production integration. We'll compare reusable models against strong dataset-specific approaches and prioritize approaches with demonstrated value.

What you'll build

• Develop prediction models over timestamped events, user attributes, transactions, product context, and relational tables.
• Experiment with structured-data and sequence architectures, cross-dataset pretraining, fine-tuning, and dataset-conditioned inference.
• Build data and training pipelines that adapt to new schemas, tasks, prediction horizons, and limited labeled histories.
• Benchmark against gradient-boosted trees, survival methods, probabilistic models, and task-specific neural networks.
• Test temporal generalization, calibration, uncertainty, transfer, compute cost, and downstream usefulness.

What success looks like

• A robust training and evaluation pipeline with strong task-specific baselines across representative business prediction problems.
• Evidence of when pretrained or adapted models improve quality, data efficiency, or deployment cost on held-out datasets.

What we look for

• Have trained and deployed models using behavioral sequences, transactions, or large tabular/relational datasets.
• Strong Python and PyTorch or JAX, including experience designing experiments and maintaining training pipelines.
• Depth in recommendation or ads ranking, sequence modeling, tabular/relational ML, or probabilistic prediction.
• Strong understanding of label leakage, temporal validation, delayed feedback, calibration, uncertainty, and distribution shift.

Relevant experience

• Tabular or relational foundation models, representation learning, meta-learning, large-scale behavioral modeling, or cross-dataset transfer.

What we offer

• Own a substantial area of research and engineering, working directly with the founders.
• Work on production ML problems and develop reusable methods and systems. Salary and equity.

Pavo is an equal opportunity employer.