AI Lab Research Roles to Watch This Week (2026)

If you are aiming for a frontier-lab move this summer, the strongest signal this week is not one generic ML job bucket. It is a cluster of roles asking for people who can turn research into measurable model improvements: RL environments, agent evaluations, Gemini pretraining data, sound generation, and sociotechnical safety work.

Status note: public job pages do not expose posting dates consistently. Where the role page shows a deadline or a "New" label, I call that out; otherwise, treat the role as a currently open posting and confirm status before applying.

Fast scan

Lab	Role / team	Location	Best fit	Status signal	Source
Anthropic	Research Engineer, Computer Use	SF / NYC / Seattle	Agentic UI use, vision/RL environments, evals	Page is marked "New"; no close date shown	posting
Anthropic	Research Engineer, RL Scaling Science	London	Long-horizon RL, scaling experiments, production training recipes	Open posting; no close date shown	posting
Anthropic	Research Engineer, Rule of Law	San Francisco	AI safety evals plus law, democracy, public-policy research	Page is marked "New"; no close date shown	posting
OpenAI	Research Engineer / Research Scientist, Personal AGI, North Stars	San Francisco hybrid	Product-facing model behavior, evals, tool use, instruction following	Open posting; no close date shown	posting
OpenAI	Research Engineer / Research Scientist, Personal AGI, Proactivity	San Francisco hybrid	Personalization, proactive agents, post-training, evals	Open posting; no close date shown	posting
OpenAI	Research Engineer / Research Scientist, RL/Reasoning	San Francisco	Frontier RL, reasoning models, alignment and capabilities	Open posting; no close date shown	posting
Google DeepMind	Research Scientist, Sound Understanding	Mountain View / Cambridge MA / New York	Audio understanding, audio-video generation, evaluation	Open posting; no close date shown	posting
Google DeepMind	Research Scientist, Pretraining, Gemini Data	Paris	Gemini pretraining for legal, finance, healthcare domains	Open posting; no close date shown	posting
Google DeepMind	Post-Doctoral Researcher, PhD, 2026 Start	Bengaluru	12-24 month AI frontier research with ML Optimization	Apply before July 31, 2026	posting

Three roles for RL and reasoning people

Anthropic - Research Engineer, RL Scaling Science. This is the most direct "scale the RL recipe" role in the batch. The team studies how RL behaves across model size, compute, and task horizon, then turns robust findings into production training recipes. The role calls for strong empirical research skills in RL or large-scale ML training, Python, and distributed systems experience; Anthropic lists the London salary range as £375,000-£640,000. 1

OpenAI - Research Engineer / Research Scientist, RL/Reasoning. OpenAI describes this team as the group driving the core reasoning paradigm behind systems such as o1 and o3. The work sits on alignment and capabilities through cutting-edge RL methods, with a San Francisco base and a listed compensation range of $295,000-$445,000 plus equity. 2

Google DeepMind - Research Engineer, Multi Agent Learning. This London role leans more engineering-heavy: building large-scale simulation platforms, research pipelines, and JAX/XLA systems for multi-agent learning. The preferred profile includes a PhD focus in ML, RL, or multi-agent systems, plus experience with large-scale training on TPUs or GPUs. 3

Two agent/product research tracks

Anthropic - Research Engineer, Computer Use. The Computer Use team is hiring for work on Claude's ability to see, use, and understand computer interfaces. Responsibilities include improving perception and agentic capabilities, building evaluation frameworks for complex computer tasks, and creating RL training environments for computer use and vision; Anthropic lists $500,000-$850,000 for the role. 4

OpenAI - Personal AGI, North Stars and Proactivity. These are two separate Research Engineer / Scientist postings inside OpenAI's Personal AGI area. North Stars focuses on model behavior, tool use, connectors, instruction following, evals, training data, and reward signals; Proactivity focuses on personalization, proactive assistants, RL, datasets, evaluations, and post-training. Both are San Francisco hybrid roles in the Research department with listed compensation of $295,000-$555,000 plus equity. 5 6

Domain and safety-adjacent research

Anthropic - Research Engineer, Rule of Law. This one is unusual for a frontier lab: the role sits in the Anthropic Institute and asks for deep AI expertise plus substantive knowledge of government, law, political science, or public policy. The work spans legal-alignment safety evaluations, fine-tuning, institutional analysis, and AI applications for civic life; the listed range is $320,000-$485,000. 7

Google DeepMind - Research Scientist, Pretraining, Gemini Data. DeepMind is looking in Paris for a Research Scientist or Research Engineer to expand Gemini pretraining models across legal, finance, and healthcare. The page asks for LLM modeling experience in pretraining or fine-tuning, ML/AI publications, and experience taking research from concept to product; the listed France range is €104,000-€107,000 plus bonus, equity, and benefits. 8

For new PhDs and publication-heavy candidates

Google DeepMind - Post-Doctoral Researcher, PhD, 2026 Start. This fixed-term role in Bengaluru is a 12-24 month research slot for frontier AI problems with Google DeepMind scientists and engineers. The posting highlights ML Optimization work in foundational models, reinforcement learning, generative modeling, causal inference, and efficiency/adaptability; applications are due before July 31, 2026. 9

Google DeepMind - Research Scientist, Sound Understanding. This is the clearest generative-audio research opening in the scan. The Sound team role covers audio understanding, transformation, generation, joint audio-video generation, audio editing, and evaluation methods for open-ended audio tasks; Google lists US pay at $147,000-$211,000 plus a 15% bonus target, equity, and benefits. 10

How I would triage applications

If your strongest evidence is shipping large ML systems, start with Anthropic Computer Use, Anthropic RL Scaling Science, and DeepMind Multi Agent Learning.
If your strongest evidence is model behavior research plus product taste, start with OpenAI Personal AGI North Stars or Proactivity.
If your strongest evidence is papers plus domain depth, look at DeepMind Pretraining Gemini Data, DeepMind Sound Understanding, and the DeepMind postdoc.
If your edge is technical AI safety plus institutions, the Anthropic Rule of Law role is narrower, but likely less crowded than generic research-engineer postings.

Before applying, check each page again for current status. Several postings have no public close date, and labs can pull roles without notice.

AI Lab Research Roles to Watch This Week