Research Scientist San Francisco or New York

Research San Francisco or New York · SF preferred Research Scientist You will define the research agenda for models that can launch and manage companies successfully. You’ll work across reasoning, post-training, continual learning, evals, and agent reliability, with the goal of making Eico’s systems better at finding demand, acquiring customers, delivering work, and turning capital into verified revenue. Research at Eico stays close to the product: the best ideas should improve the agents, the datasets, and the market-like evals we are known for. We are hiring a small number of unusually strong technical people early, and this role should help set the research and technical bar for the company. What you’ll own: 01 Research directions across agent capability, post-training, continual learning, evaluation, and company-building performance. 02 Datasets built from real company-building traces: demand discovery, customer acquisition, delivery, support, revenue, failure, and verification. 03 Training, post-training and continual learning of Eico models in house: fine-tuning, SFT, RL, reward modeling, preference data, synthetic data, tool-use learning, and related methods. 04 Market-like evals that test whether agents can find something worth doing, acquire a customer, deliver value, recover from failure, and produce verified outcomes. 05 Agent backends and harnesses for the product, working closely with AI engineers, software engineers, and incubation teams. 06 Public-facing research: papers, blog posts, technical reports, Twitter / X, podcasts, conferences, and the ideas Eico is known for. What we’re looking for: PhD required in computer science, machine learning, mathematics, statistics, or a related technical field. First-author or otherwise major publications at top venues such as NeurIPS, ICML, ICLR, EMNLP, COLM, or equivalent. At least 5 years of experience, or equivalent depth demonstrated through exceptional research and shipped systems. AI-lab, model-company, or serious research experience: OpenAI, Anthropic, DeepMind, FAIR / Meta GenAI, Thinking Machines, SSI, Tesla AI, xAI, Mistral, Cohere, Character, Adept, Cursor, Cognition, Harvey, Ramp AI, Glean AI, Perplexity, Replit AI, Databricks / Mosaic, Together, Hugging Face, Physical Intelligence, or equivalent. Has personally trained, fine-tuned, post-trained, and materially improved large models. Has built agent systems, model infra, evals, data pipelines, RAG, tool-use systems, inference routing, or training infrastructure. Strong researcher: reads deeply, has taste, forms opinions, chooses important problems, and runs rigorous experiments. Strong engineer: can ship production systems, not just notebooks or papers. Strong empirical judgment: knows the difference between a real result, noise, leakage, and a bug. Public technical credibility: can write, speak, publish, and represent Eico to serious AI people. High-agency operator. Low ego, high standards, fast execution. This is probably not for you if: Your AI experience is mostly prompt engineering or API wrappers. You have only built AI features, but have not gone deep on training, post-training, evals, or model behavior. You are purely academic and do not want to ship. You are purely backend/product and weak on ML research. You do not want to be a public technical face. #J-18808-Ljbffr Economic Intelligence Co.