Position Overview
Sentra is building organizational superintelligence through memory infrastructure that reasons across time, causality, and context. As a Research Scientist, you will tackle fundamental problems in knowledge representation, temporal reasoning, and semantic compression. You will design and implement systems that maintain execution state for entire organizations, consolidate millions of micro-events into durable knowledge, and learn patterns that predict events before it happens.
Key Responsibilities
- Build LLM-powered information extraction pipelines that process unstructured communications and text data into structured entity-relationship representations.
- Develop memory consolidation algorithms that validate information through multiple observations, merge duplicate entities, and prune ephemeral data.
- Design temporal knowledge graph architectures that model organizational execution state as living, continuously updated systems rather than static records.
- Create graph attention mechanisms and reasoning systems for complex causal queries about blockers, dependencies, and outcome patterns.
- Research lossy semantic compression using information-theoretic principles to condense event streams into query-relevant long-term memory.
- Design entity resolution systems handling identity evolution where entities merge, split, and transform through time.
- Build meta-learning systems that identify organizational patterns and recognize when current situations match historical success or failure indicators.
- Develop privacy-preserving cross-organizational learning using federated learning and differential privacy techniques.
- Publish research findings and contribute to the broader research community on knowledge graphs and organizational intelligence.
Must-have Requirements
- 5+ years building novel systems in machine learning, NLP, knowledge graphs, or related areas with evidence through publications, production implementations, or significant open-source contributions.
- Deep knowledge of knowledge graphs, graph neural networks, or temporal reasoning demonstrated through shipped systems and architectural exploration.
- Strong ML and NLP foundation, particularly in information extraction, entity resolution, or semantic representation.
- Proficiency in Python and modern ML frameworks (PyTorch preferred) with experience deploying models at scale.
- Track record of publishing research (conference papers, technical blog posts, or detailed technical documentation) and exploring novel architectures.
- Ability to move between theoretical investigation and practical implementation, shipping research into production.
Bonus skills
- Graph databases (Neo4j, TigerGraph, Neptune) and query optimization for large-scale graphs.
- Information theory, compression, or temporal data structures.
- Causal inference, probabilistic reasoning, or Bayesian methods.
- Distributed systems, stream processing, or real-time ML serving.
- Human memory and cognition models.
- Privacy-preserving ML (federated learning, differential privacy, secure multi-party computation).
- Enterprise AI systems, workflow automation, or organizational software.
- Publications at top-tier conferences (NeurIPS, ICML, ICLR, KDD, EMNLP, ACL, WWW, SOSP, OSDI).