Research Product Manager – AI Systems

About Granica

Granica is building the efficiency and intelligence layer for enterprise AI
.

• Crunch makes massive enterprise data cheaper and easier to operate.
• Large Tabular Models learn from structured data to support shared intelligence across many capabilities.
• Myelin makes long-running AI agents more efficient and durable.

Granica has processed hundreds of petabytes of tabular data in production
, and our research is led by Stanford Professor Andrea Montanari
.

• Location: Mountain View, CA
• Work model: On-site, five days per week
• Level: Senior / Staff / Principal

Granica is hiring a Research Product Manager to turn frontier AI research into systems that create real value from enterprise data.

You’ll work at the intersection of AI/ML systems, structured data, research, and product
, helping define:

• how models learn from real-world data
• how model quality and emerging capabilities are evaluated
• how research becomes production systems
• how technical improvements translate into economic value

Experience with structured or tabular data is a major advantage, but we are equally interested in exceptional product leaders from AI systems, ML infrastructure, evaluation, training/post-training, and applied ML
.

This is not a traditional feature PM role. You’ll work directly with researchers and engineers to turn technically ambitious ideas into products and systems.

Most valuable enterprise data is structured, relational, private, and constantly changing
.

Today, companies typically build machine learning one problem at a time: define a target, prepare data, train a model, deploy it, and repeat for the next problem.

Granica’s research is pioneering a fundamentally better approach.

We are building models that learn the underlying structure and distributions of enterprise data deeply enough that shared intelligence can support many capabilities — including prediction, anomaly detection, classification, forecasting, imputation, synthetic data, and risk modeling.

The goal is to move beyond one model per task
.

• Define product direction for AI systems that learn from structured and relational data
• Partner with researchers to translate new model capabilities into production systems
• Define how model quality and emerging capabilities are evaluated
• Identify enterprise ML problems that can move from task-specific models toward shared intelligence
• Connect AI systems with enterprise data platforms, warehouses, and lake houses
• Translate model improvements into measurable customer and economic value
• Drive research from experiment → system → product → customer value
• Shape the roadmap around the highest-value enterprise problems

Structured enterprise data is fundamentally different from natural-language corpora.

Models must understand:

• schemas and metadata
• joins and relationships
• heterogeneous data types
• distributions and missingness
• temporal behavior
• business-specific context

The goal is to build models that understand enterprise data deeply enough that many useful capabilities emerge from the same underlying intelligence.

A benchmark score alone cannot tell us whether a model has truly learned the structure of enterprise data.

We care about:

• whether capabilities are reliable
• how uncertainty is measured
• which improvements generalize
• when research is production-ready
• when better model performance creates real economic value

Evaluation is part of the product and research system itself.

• 5+ years of product leadership or equivalent technical ownership in AI/ML, data systems, infrastructure, or applied research
• Strong technical judgment and ability to work directly with researchers and engineers
• Experience taking complex…