Fits

Office

Location:


Wellesley, MA (3 days per week in the office)

Accrete is an agentic managed services company: we deliver the AI infrastructure and the AI workforce that does high-stakes, judgement-heavy work for government and enterprise clients — at the economics of software, not labor.

At the core is Accrete’s Knowledge Engine, a dynamic context graph that captures an organization’s tacit knowledge, resolves data across silos, and builds a living, auditable ground truth. Expert agents reason against that ground truth to act on complex, high-stakes decisions — not just answer questions about them. From national security to commercial operations, Accrete delivers the work on one platform, with unlimited expert agents and expert judgement.

We are looking for a Senior Applied Scientist to contribute to the research and development behind Accrete's context graphs — the intelligence layer the Knowledge Engine and its agents are built on.

The work draws on knowledge representation, graph-based machine learning and information retrieval, combined with LLMs
, to continuously construct, enrich, reason over, and maintain a structured picture of how an organization operates.
Agentic systems are key: the graph and the agent topology built on top of it are designed together, since what agents need to retrieve and act on shapes how the graph is structured.

The central challenge is to move past static knowledge graphs and conventional retrieval toward dynamic context graphs that capture how decisions get made and why — the actors, events, risks, and rules involved, the dependencies among them, and how all of it shifts over time. Much of what matters was never written down directly; it has to be reconstructed from the traces work leaves in tickets, threads, and meetings.

In practice that means extraction from large volumes of heterogeneous data, resolving entities and relationships across sources, modeling provenance and uncertainty explicitly, and building graph structures agents can reason over reliably.

This is an applied research role. Ideas become systems that run on real data at scale, developed alongside engineers and product teams. It also means taking evaluation seriously: there is no easy ground truth for whether a graph captured the right reasoning, and designing the benchmarks and adjudication methods that answer that question is part of the work, not an afterthought.

• Conduct applied research focused on context graphs, knowledge representation, graph intelligence, and AI systems operating over complex, heterogeneous data.

• Develop algorithms and architectures for constructing and continuously updating dynamic context graphs from structured and unstructured data.
• Research methods for entity resolution, relationship extraction, event extraction, temporal reasoning, semantic linking, and knowledge discovery across disparate information sources.
• Develop graph-based representations of entities, relationships, events, decisions, processes, and organizational knowledge.
• Explore approaches for incorporating time, provenance, confidence, uncertainty, and source attribution into graph-based representations.
• Develop graph-based methods for supporting LLM reasoning, retrieval, planning, and agentic workflows.
• Research techniques for combining LLMs with structured graph representations to improve reasoning accuracy, grounding, and explainability.
• Design and evaluate graph algorithms, embeddings, retrieval methods, and machine learning approaches for discovering latent relationships and relevant context.
• Investigate methods for identifying changes, inconsistencies, gaps, and emerging patterns within evolving knowledge and context graphs.
• Develop prototypes and…