Research engineer vs scientist
- A career coach outlined differences: Research Engineers scale ideas; Research Scientists focus on discovery. - He noted research scientist roles often require a PhD and have <0.5% acceptance, while RE hiring is 2–5× easier. - The post highlights distinct hiring bars, pay gaps, and decision frameworks for candidates choosing which track to pursue. (x.com)
At frontier artificial intelligence labs, the split between research engineer and research scientist is less about coding skill than about who decides the next idea. (sundeepteki.org) Sundeep Teki, an artificial intelligence career coach and former Amazon Alexa research scientist, wrote on April 19 that research engineers are hired to make ideas work at scale, while research scientists are hired to decide what the lab should work on next. His post was aimed at candidates targeting OpenAI, Anthropic, and Google DeepMind. (sundeepteki.org) (interviewquery.com) In practical terms, that means the engineer role centers on training runs, infrastructure, optimization, and implementation, while the scientist role centers on new algorithms, experiments, and research direction. OpenAI’s current job pages reflect that split: its research engineer posting emphasizes building systems, and its research scientist posting says the role advances the team’s research agenda. (openai.com 1) (openai.com 2) The hiring bar is also different. Teki wrote that research scientist pipelines at frontier labs often have acceptance rates below 0.5% and usually favor candidates with a PhD and a publication record, while research engineer hiring is typically two to five times easier for strong candidates with large-scale machine learning and systems experience. (sundeepteki.org 1) (sundeepteki.org 2) Pay follows the same divide. Levels.fyi data cited by Teki shows OpenAI research scientist compensation in the United States ranging from $771,000 to $1.47 million a year, while his April 19 guide says OpenAI research engineers top out around $530,000, leaving a senior-end gap of more than $900,000. (levels.fyi) (sundeepteki.org) That distinction has become more visible as labs blur titles on some teams. OpenAI has active postings labeled “Research Engineer/Research Scientist” for reinforcement learning and reasoning work, and Anthropic has listed combined roles that ask for both advanced degrees and strong software engineering. (openai.com) (jobs.accel.com) The mixed postings do not erase the underlying split. Anthropic’s careers page says it is building a safety-first research company, and DeepMind’s careers page says its teams include both scientists and engineers working together on the same systems, which helps explain why candidates often see overlapping responsibilities with different evaluation standards. (anthropic.com) (deepmind.google) Teki’s advice is to choose the track by evidence, not aspiration: publications, original research taste, and problem selection point toward scientist roles, while systems work, scaling skill, and fast implementation point toward engineer roles. He wrote that many candidates lose months preparing for the wrong loop because they mistake the engineer track for a lighter version of scientist hiring. (sundeepteki.org) The result is a labor market where two people can sit on the same model team, contribute to the same paper, and still be hired, paid, and promoted on different logic. For candidates trying to break into frontier labs in 2026, that title now functions less like a label and more like a career bet. (sundeepteki.org)
