Applied Scientist

Pavo is building Enterprise Superintelligence: compounding systems that take ownership of business outcomes and work with humans to deliver them. We believe that while foundation models are necessary, they are not sufficient. The hard problem is systems intelligence: end-to-end architectures that understand a company’s code, data, and decisions, and improve themselves through experience. We are assembling a small, senior team of researchers and engineers obsessed with systems-first intelligence. Our current team consists of PhDs and ML engineers from top applied ML and coding agent companies, with a heritage of shipping systems at Spotify, ShareChat, and Sourcegraph scale. Our team has built impressive momentum with a small group of highly capable engineers and researchers. 02 The role As an Applied Scientist at Pavo, you will lead the science track of tribal-knowledge generation. You’ll work on the open problems that sit between today’s RAG and tomorrow’s organizationally-aware agents, and turn them into shipped, evidence-backed improvements to the production system. This is applied research in the truest sense: the questions arise from real production behavior, the answers must improve it, and the cycle from interesting finding to shipped change is days, not quarters. The questions themselves are also publishable, most sit at or beyond the current literature. This is a senior, individual-contributor role. Everyone on the team joins as a Member of Technical Staff, with the scope, autonomy, and end-to-end ownership that title implies. We are looking for an applied researcher who turns messy production behavior into questions, and questions into shipped, evidence-backed change. 03 What you’ll work on Retrieval over Heterogeneous Private Evidence: How an agent should traverse an organization’s source code, structured data, internal documents, and conversations to assemble the evidence required to compile knowledge. Verifiability of Open-Ended Generation: What it means for an agent-produced knowledge artifact to be trustworthy, beyond precision-only validation of individual facts. Evaluation of Multi-Stage Agentic Pipelines: Benchmarks and instrumentation that localize quality gains to the responsible stage, without leaking the answer key into the pipeline being measured. Reliability & Variance: Characterizing and reducing run-to-run variance in stochastic synthesis, so knowledge artifacts can be released with the same confidence as deterministic software. Continual Update & Conflict Resolution: How a compiled knowledge artifact should evolve as the underlying organization changes, surfacing conflict and accruing authority and temporal validity. Publication: Internal findings as decision-grade memos; external results as papers, talks, or technical reports, wherever the work advances the field. 04 What we look for Senior Track Record: Years of applied-research or ML experience (typically 8+ in industry, or a PhD plus a strong applied-research record), including work you drove end-to-end that held up under scrutiny, the scientist others bring their hardest, most ambiguous problems to. Working Understanding of Agentic Systems: You know how tool use, multi-turn execution, context limits, and structured outputs behave in practice, even if you haven’t built a production agent yourself. Strong Retrieval Fundamentals: Fluency in dense and sparse retrieval, reranking, query understanding, and IR-style evaluation. Many of the open problems here are dressed-up retrieval problems. Experimental Discipline: You’ve designed and run ablations that survive scrutiny; you treat n=1 with the suspicion it deserves; you know the difference between a result that explains the past and one that predicts the future. Familiarity with the Hallucination & RAG-Eval Literature: At a level where you can identify when a published benchmark or method has structural limitations. Production Intuition: You can read messy run logs and formulate the question hiding inside them. Strong Technical Writing: You can produce a finding another scientist trusts, and a script the engineering team can run. 05 Nice to have Publications in agents, RAG, IR, hallucination evaluation, knowledge integration, or continual learning. Hands-on experience designing benchmarks or evaluation harnesses for open-ended generation. Familiarity with conflict-resolution, record-linkage, or entity-resolution literature, these surface as adjacent problems in tribal knowledge. PhD in ML / NLP / IR, or an equivalent applied-research track record in industry. 06 What we offer Foundational Work: The private knowledge layer will reshape how AI agents operate inside organizations. The problems are real and at the edge of the field. Short Loop: Work directly with the engineering lead and the founders. Finding to recommendation to shipped change is days, not quarters. Real Ownership of the Science Agenda: In a small, technically deep team. Your name will be on the work. Publication Encouraged: Including external, papers, talks, and technical reports where the work advances the field. Pavo is an equal opportunity employer. We celebrate diversity and are committed to creating an inclusive environment for all employees. #J-18808-Ljbffr