Interested in this Research Engineer role at Cognition AI?
Apply Now →Skills & Technologies
About This Role
Location
San Francisco
Employment Type
Full time
Location Type
On-site
Department
Research & DevelopmentResearch
We are an applied AI lab building end-to-end software agents.
We're the makers of Devin, the first AI software engineer.
Our team is extremely talent-dense. Among our founding team, we have world-class competitive programmers, former founders, and leaders from companies at the cutting edge of AI including Scale AI, Palantir, Cursor, Waymo, Tesla, Lunchclub, Modal, Google DeepMind, and Nuro.
Building Devin is just the first step—our hardest challenges still lie ahead. If you’re excited to solve some of the world’s biggest problems and build AI that can reason on real-world tasks, apply to join us.
Role Mission
Devin writes and runs code, uses tools, takes actions in real customer systems, and operates for hours without a human in the loop. Safety here is not a research paper; it is whether an agent behaves correctly in production, millions of times a day. You will be a founding member of Cognition's safety team, reporting to the Head of Safety. You will build the evaluations and red-teaming that gate every release, develop alignment methods that shape how our models are trained, and work directly with the agent team on how Devin plans, acts, and asks for help. This is a hands-on research engineering role for someone who wants their safety work to run in the loop of a real agent, not sit in a benchmark.
What You'll Accomplish
- Build agentic safety evals: Design and run evaluations for the failure modes that matter for autonomous agents: unsafe actions, prompt injection, data exfiltration, sandbox escape, reward hacking, and misuse. Make them fast enough to run on every model and product release.
- Red-team Devin: Attack the agent and its harness systematically. Find the failures before customers do and turn them into fixes and regression tests.
- Develop alignment methods: Work with post-training on reward modeling, preference data, constitutional approaches, and other techniques that make the model safer without making it worse at the job.
- Shape the agent harness: Partner with the agent team on permissions, oversight, and escalation: when Devin should act, when it should ask, and how it should explain itself.
- Publish and share: Contribute to Cognition's external safety work through papers, evals, and open methods where it makes sense.
Exceptional Candidates Have Demonstrated
- Safety or alignment research experience: Hands-on work on evaluations, red-teaming, alignment, or interpretability at a frontier AI lab or in published research. You have built things that changed how a model was trained or deployed.
- Agentic systems understanding: You know how agents fail in practice: tool misuse, specification gaming, long-horizon drift, adversarial inputs. You have measured it or have concrete ideas about how to.
- Strong engineering fundamentals: Proficiency in Python and PyTorch (or JAX). You can build eval infrastructure, run experiments at scale, and read the training and inference code.
- Empirical rigor: You design clean experiments, report results honestly, and know the difference between a real improvement and noise.
- Comfort with adversarial thinking: You enjoy breaking systems and are good at it.
- Relevant industry experience: Prior experience at a frontier AI lab, applied AI company, or developer tools company.
- Advanced degree: PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline; or equivalent industry research experience.
Equal Opportunity
Cognition is an equal opportunity employer. We do not discriminate on the basis of race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, veteran status, or any other protected characteristic under applicable law. We are committed to providing reasonable accommodations for candidates with disabilities throughout the hiring process - please let us know if you need any.
Role Details
About This Role
Research Engineers bridge the gap between research and production. They implement papers, build experiment infrastructure, optimize training pipelines, and make research prototypes production-ready. They're the engineers who make research work at scale.
The role sits at a unique intersection. You need to understand the math well enough to implement novel architectures correctly, and you need the engineering chops to make them run efficiently on distributed systems. When a research scientist has a breakthrough idea, you're the person who turns it from a notebook prototype into a training pipeline that runs on 256 GPUs.
Across the 9,079 AI roles we're tracking, Research Engineer positions make up 2% of the market. At Cognition AI, this role fits into their broader AI and engineering organization.
Research Engineer roles are growing as AI labs recognize that research velocity depends on engineering quality. The role is less competitive than Research Scientist (no PhD required), but the bar for engineering skill is very high. These roles are concentrated at major labs and well-funded startups.
What the Work Looks Like
A typical week involves: implementing a new attention mechanism from a recent paper, profiling and optimizing a training pipeline that's bottlenecked on data loading, building evaluation infrastructure for a new benchmark, debugging distributed training issues across a GPU cluster, and pair-programming with a research scientist on their latest experiment. The work is deeply technical.
Research Engineer roles are growing as AI labs recognize that research velocity depends on engineering quality. The role is less competitive than Research Scientist (no PhD required), but the bar for engineering skill is very high. These roles are concentrated at major labs and well-funded startups.
Skills Required
Strong software engineering fundamentals plus ML knowledge. Python, C++, and CUDA experience are common requirements. You'll need to read papers and turn ideas into working code. Distributed systems experience (especially distributed training) is highly valued. Performance optimization skills separate great candidates from good ones.
Experience with large-scale training infrastructure (FSDP, DeepSpeed, Megatron), GPU programming (CUDA, Triton), and the internals of ML frameworks (PyTorch internals, custom autograd functions) is what makes candidates stand out. The best research engineers can debug issues that span the full stack from GPU memory management to numerical precision to algorithmic correctness.
Strong postings mention the team's recent research, the infrastructure scale, and the specific technical challenges. They often list the research areas you'd support. Look for roles that emphasize both implementation quality and research understanding.
Compensation Benchmarks
Research Engineer roles pay a median of $252,000 based on 457 positions with disclosed compensation. Mid-level AI roles across all categories have a median of $193,200.
Across all AI roles, the market median is $213,000. Top-quartile compensation starts at $263,911. The 90th percentile reaches $318,400. For comparison, the highest-paying categories include AI Safety ($256,000) and AI Engineering Manager ($251,622). By seniority level: Entry: $111,000; Mid: $193,200; Senior: $229,300; Director: $272,300; VP: $235,830.
Cognition AI AI Hiring
Cognition AI has 1 open AI role right now. They're hiring across Research Engineer. Based in San Francisco, CA, US.
Location Context
AI roles in San Francisco pay a median of $261,000 across 2,646 tracked positions. That's 23% above the national median.
Career Path
Common paths into Research Engineer roles include Software Engineer, ML Engineer, Research Intern.
From here, career progression typically leads toward Senior Research Engineer, Research Scientist, ML Architect.
This is one of the best entry points into AI research without a PhD. Build a strong engineering portfolio with ML projects, contribute to open-source ML frameworks, and demonstrate that you can implement complex ideas correctly and efficiently. The transition to Research Scientist is possible with published first-author work, which some research engineer roles support.
What to Expect in Interviews
Technical screens test both engineering skill and research understanding. Expect coding rounds with performance-critical implementations (GPU optimization, efficient data loading). Be prepared to discuss papers relevant to the team's research area and explain how you'd implement key ideas. System design questions focus on training infrastructure: distributed training, experiment tracking, and compute resource management.
When evaluating opportunities: Strong postings mention the team's recent research, the infrastructure scale, and the specific technical challenges. They often list the research areas you'd support. Look for roles that emphasize both implementation quality and research understanding.
AI Hiring Overview
The AI job market has 9,079 open positions tracked in our dataset. By seniority: 106 entry-level, 4,384 mid-level, 3,343 senior, and 1,246 leadership roles (Director, VP, C-Level). Remote roles make up 6% of the market (531 positions). The remaining 8,478 roles require on-site or hybrid attendance.
The market median for AI roles is $213,000. Top-quartile compensation starts at $263,911. The 90th percentile reaches $318,400. Highest-paying categories: AI Safety ($256,000 median, 66 roles); Research Engineer ($252,000 median, 457 roles); AI Engineering Manager ($251,622 median, 55 roles).
Research Engineer roles are growing as AI labs recognize that research velocity depends on engineering quality. The role is less competitive than Research Scientist (no PhD required), but the bar for engineering skill is very high. These roles are concentrated at major labs and well-funded startups.
The AI Job Market Today
The AI job market spans 9,079 open positions across 16 role categories. The largest categories by volume: AI/ML Engineer (6,862), AI Product Manager (453), AI Software Engineer (451). These three account for the majority of open positions, though smaller categories often have higher per-role compensation because of specialized skill requirements.
The seniority mix tells a story about where AI teams are in their maturity. Entry-level roles (106) are outnumbered by mid-level (4,384) and senior (3,343) positions, reflecting that most companies are past the 'build a team from scratch' phase and need experienced engineers who can ship production systems. Leadership roles (Director, VP, C-Level) total 1,246 positions, representing the bottleneck between technical execution and organizational strategy.
Remote work availability sits at 6% of all AI roles (531 positions), with 8,478 requiring on-site or hybrid attendance. The remote share has stabilized after the post-pandemic correction. Senior and specialized roles (Research Scientist, ML Architect) are more likely to be remote-eligible than entry-level positions, partly because experienced hires have more negotiating power and partly because these roles require less hands-on mentorship.
AI compensation is structured in clear tiers. The market median sits at $213,000. Top-quartile roles start at $263,911, and the 90th percentile reaches $318,400. These figures include base salary with disclosed compensation. Total compensation (including equity, bonuses, and sign-on) runs 20-40% higher at companies that offer those components.
Category matters for compensation. AI Safety roles lead at $256,000 median, while Prompt Engineer roles sit at $141,440. The spread between highest and lowest-paying categories reflects the premium on specialized technical skills versus broader analytical roles.
The most in-demand skills across all AI postings: Python (1,649 postings), Aws (898 postings), Azure (723 postings), Rag (671 postings), Gcp (536 postings), Pytorch (434 postings), Claude (410 postings), Kubernetes (406 postings). Python dominates, appearing in the vast majority of role descriptions regardless of category. Cloud platform experience (AWS, GCP, Azure) is the second most common requirement. The newer entrants to the top skills list (RAG, vector databases, LLM APIs) reflect the shift from traditional ML toward generative AI applications.
AI Hiring Overview
The AI job market has 9,079 open positions tracked in our dataset. By seniority: 106 entry-level, 4,384 mid-level, 3,343 senior, and 1,246 leadership roles (Director, VP, C-Level). Remote roles make up 6% of the market (531 positions). The remaining 8,478 roles require on-site or hybrid attendance.
The market median for AI roles is $213,000. Top-quartile compensation starts at $263,911. The 90th percentile reaches $318,400. Highest-paying categories: AI Safety ($256,000 median, 66 roles); Research Engineer ($252,000 median, 457 roles); AI Engineering Manager ($251,622 median, 55 roles).
Frequently Asked Questions
Get Weekly AI Career Intelligence
Salary data, skills demand, and market signals from 16,000+ AI job postings. Every Monday.