You will shape how AI agents are evaluated in immunology by contributing to the technical approach and benchmark design.
You will build ground-truth datasets from studies, your own experience, and novel work to test expert-level reasoning.
You will collaborate with software engineers and computational biologists to rigorously measure agent performance on real biological tasks.
Latch builds rigorous scientific benchmarks for AI agents in biology, working with frontier labs and pharma to measure real biological workflows. The culture emphasizes radical transparency, heavy AI use, and original expert judgment, with a collaborative team that enjoys outings, sports, and game nights.
Serve as the scientific authority on ground truth for AI agents in neuroscience, testing expert-level reasoning across all system levels.
Build benchmark tasks and datasets for normal function and disease, curating multi-modal data and writing gold-standard expert answers.
Set hard calls on contested claims and interrogate agent reasoning, challenging fluent but unsound conclusions.
Latch builds rigorous scientific benchmarks for AI agents across biology, working with frontier labs and pharma to test whether agents can handle real biological workflows. The culture is radically transparent, opinionated but respectful, with structured autonomy, regular outings, and a preference for onsite work in SF.
Contribute to the technical approach for evaluating AI agents in diagnosing, treating, and researching rare and progressive diseases.
Build ground-truth benchmark datasets from studies and novel work that rigorously test agents' analytical and interpretive skills.
Collaborate with scientists, physician-scientists, engineers, and computational biologists to define strategy across multiple benchmarks.
Latch builds rigorous scientific benchmarks for AI agents across biology, measuring whether agents can handle real biological workflows. An early, tight-knit team, Latch practices radical transparency, uses AI constantly, and fosters a culture of respect, expert judgment, and community activities.
Contribute to the technical approach for evaluating AI agent performance across the genetic medicines pipeline, including gene therapy, gene editing, and RNA therapeutics.
Design realistic program case studies spanning target validation, drug design, screening, chemistry optimization, and IND-enabling studies.
Review and stress-test AI agent outputs for scientific soundness, calibrating confidence on preclinical data with incomplete or conflicting signals.
Latch builds rigorous scientific benchmarks for AI agents across biology, working with frontier labs and pharma to measure whether agents can handle real biological workflows. The culture emphasizes radical transparency, high agency, and using AI tools with original expert judgment.
Evaluate AI-generated responses for scientific accuracy and logical consistency.
Test AI models with theoretical biology questions and realistic laboratory scenarios.
Identify and document model failure modes, recommending improvements to prompts and evaluation methods.
A partner company of Jobgether focused on improving AI language models through expert scientific training and evaluation. The organization works with freelance specialists to enhance AI reasoning and reliability in biological domains.
Set the scientific ground truth for TxBench, building file-based tasks and datasets across therapeutic antibody development workflows.
Test whether AI agents can distinguish real quality-attribute failures from analytical artifacts across CMC, PK, immunogenicity, and safety.
Interrogate agent reasoning step by step, score confidence, and make expert go/no-go calls that hold agents to development-review standards.
Latch builds rigorous scientific benchmarks for AI agents across biology, working with frontier labs and pharma. The culture is radically transparent, AI-driven, and rewards original expert judgment, with regular team outings and a preference for onsite collaboration.
Define expert reasoning for peptide discovery and write gold-standard answers with defended rationale.
Curate datasets from real programs and build Python analysis pipelines that produce verifiable ground truth.
Test and challenge AI agents' reasoning, scoring confidence and pinpointing where they go wrong.
Latch builds rigorous scientific benchmarks for AI agents across biology, working with frontier labs and pharma to measure real biological workflows. It's a transparent startup that uses AI constantly and rewards original expert judgment.
Build evaluations to measure and improve AI reasoning in complex, high-stakes decisions.
Collaborate with engineers, product, and domain experts to turn observed failures into testable improvements.
Own the quality of shipped systems, from code reviews to monitoring and integration.
Elicit builds AI systems that help research and business teams make evidence-based decisions on high-stakes questions. They are a remote-friendly, collaborative team that values rigorous thinking, experimentation, and in-person retreats.
Define gold-standard answers for AI agents from raw neutralization, epitope, and drug-response data.
Build tasks testing whether agents separate real escape or resistance from assay noise.
Judge agent output with expert-level calls in vaccine, antibody, and antiviral development.
Latch builds rigorous scientific benchmarks for AI agents across biology, working with frontier labs and pharma. The company practices radical transparency and a collaborative culture, with regular outings and off-sites.
Build and post-train agents for scientific research, including literature search, code execution, and domain tool use.
Develop RL and multi-agent methods for proposing, testing, and critiquing hypotheses, with strong baselines and evaluation.
Design verifiers, reward signals, and reproducible experimental loops for scientific workflows, ensuring evidence-based improvements.
Prime Intellect builds the open superintelligence stack, providing infrastructure for frontier AI labs to train, evaluate, and deploy models at scale. The company has raised $150M from top investors and operators, and is building a culture focused on frontier research, real infrastructure, and go-to-market for a new category.
Own inbox triage, calendar management, travel, and AI briefings.
Work on a 5+3 split shift with live overlap in PST and async prep.
Use AI tools like Claude and ChatGPT to research and manage tasks.
Assist World is a remote-first company that provides virtual assistant staffing solutions to innovative businesses. They foster a culture of autonomy and flexibility, with a 100% remote team and no tracking.
Develop AI methods and algorithms for reliable agents.
Design simulations and evaluations for agent behavior.
Partner with product and engineering to ship research.
The company builds an AI agent platform focused on making agents reliable in production. It is a research-driven environment with a collaborative, startup-oriented culture.