AI can now operate a computer about as well as a person. Knowing what that person is trying to do, and when to stay quiet, is still unsolved. That's the problem you'll own.About Lucy Labs Lucy is a proactive AI coach from Lucy Labs that lives on your desktop. With each person's permission she understands how they work, asks before she answers, and coaches them to build the AI solutions their work and life need. You decide, design and direct; Lucy coaches; whatyou build does the work. We're early, working with a small group of design partners. Today Lucy coaches when people ask; proactive coaching is planned, and this role helps make it possible.The role You'll develop methods that turn incomplete observations of work into useful, correctable interpretations. Someone researching a company might be preparing for a meeting, assessing a supplier or simply learning. You'll help Lucy tell those apart, ask for clarification when it helps and stay quiet when the evidence is weak, without pretending an observed action reveals someone’s intention.What you'll do Develop and compare methods for Action-to-Task-to-Goal inference from contextual and temporal evidence. Start with inspectable rules and baselines, and earn complexity through measured improvement.Find recurring work patterns and rank opportunities for workflows or solutions, keeping observed sequences, candidate interpretations and user-confirmed goals separate.Define confidence, calibration, abstention and correction behavior: when more information improves a recommendation, and when it only adds cost, delay or intrusion.Build reproducible experiments, error analyses and prototypes that production engineers can implement, with their limits documented.Connect methods to our knowledge graph of tasks, tools,



skills and prerequisites.Handle drift, cold starts, ambiguous vocabulary and feedback bias within permission, minimization and retention requirements.Who you'll work with AI Engineering productionizes your methods and AI Evaluation tests them independently. Knowledge Graph Engineering, Behavioral Science, Product Data Science and Product Management are your closest partners. You own the inference method, its uncertainty and its limits.What you bring Quantitative methods applied to noisy event, sequence, language or behavioral data, including a failed experiment you learned from.Strong Python, statistics and reproducible experimentation.Helpful: ranking, probabilistic reasoning, information retrieval or interpretable machine learning.Evidence of turning research into a working product. We value that more than a particular degree or publication count.Your first 90 days Pick one narrow inference problem, establish its simplest credible baseline, build an error taxonomy with AI Evaluation and improve one material failure mode against independent examples. Then hand engineering a reproducible prototype with explicit constraints, including what remains unproven.Why Lucy Labs Trust is the architecture: Lucy works for the person, even though the company pays for her, and our terms prohibit using Lucy for employee surveillance, ranking or disciplinary action.Ground-floor impact and real ownership, with equity that reflects contribution and risk.An AI-native team: Lucy Labs runs on Lucy, and the skills you build here travel with you.Remote-first and asynchronous, with R&D based in Spain.Pay €80,000 to €110,000 gross base salary per year, plus equity.Lucy Labs welcomes applicants of every background.#J-18808-Ljbffr

📌 Applied Scientist, Behavioral Ai (Valencia)
🏢 Lucy Labs
📍 Valencia