AI Knowledge Architect (Scientific Intent)

Albert Invent•Oakland, CA

10d•Hybrid

About The Position

We are looking for somebody who is good at deciding how scientific knowledge should be represented. A Knowledge Architect (Scientific Intent) to join our growing AI/ML team. This role has a large hand in how scientific knowledge is modeled, captured, and reasoned over inside Albert OS. Albert's data model captures what happened in the lab: materials, quantities, process parameters, transformation stages, and how they connect to inventory and regulatory context. The reasoning behind that work lives in a different world, in narrative and unstructured context: why a formulation was revised, what a chemist ruled out and why, the thinking patterns that shape a process over months. Connecting those worlds is where scientific reasoning becomes computable, and it is the problem this role takes on. You will define the ontologies Albert needs to represent that, decide how the data model should be extended to hold it, and work with engineers to make those extensions real.

Requirements

Motivated by the problem of how scientific knowledge gets captured, and want defining that to be your job rather than a side effect of it.
Can walk through a system you modeled and speak sharply to the tradeoffs: what you made a first-class entity versus an attribute, how you handled schema and versioning over time, how you reasoned about cardinality and composition. Formal knowledge-representation vocabulary is not required going in.
Work comfortably at the schema and data-model level with engineers, with enough fluency in graph structure and typed relationships to co-design credibly.
Have a scientific, industrial, or lab-adjacent background, or enough curiosity about experimental science to get fluent in it fast.
Have a formal ontology, taxonomy, or knowledge-representation background, or you want to build one. We will support ramping into it, including close collaboration with the VP, AI/ML on foundational ontology design.
Applied ML, numerical modeling, or large-scale data science experience is useful here, since the structure you define is what downstream models and retrieval depend on.
Push back on adding a free-text field when the harder structural work is what the problem actually needs.

Responsibilities

Define the ontologies and taxonomies Albert needs for scientific intent and process narrative, including the types, relationships, and attachment points that connect reasoning to structured experimental data.
Decide how the core data model should be extended, and own the review that keeps new structure traversable, reasoning-friendly, and AI-native rather than accumulating as a dumping ground.
Figure out how to capture scientific intent as chemists work, including what is worth noting as they use the platform and where annotation can enrich the knowledge representation already in the system.
Partner closely with engineers on sophisticated implementations of these data models and ontologies, from typed relationships and versioning through to how the structure behaves in production.
Work with process development chemists to understand how they actually reason through experiments, and with AI/ML engineers building on the graph, so the structure serves real retrieval and reasoning needs.
Create the playbooks and review practices that keep the knowledge layer coherent as more teams write to it, and help grow this discipline as the team expands.