Research scientist is the role that defines the frontier of a discipline. In ML and AI, research scientists design experiments, push novel methods, publish at top venues, and shape where the field goes next. The role is distinct from applied scientist (which specialises existing research to product) and from ML engineer (which builds the systems that serve models in production). The frontier AI labs have made research scientist one of the most coveted and compensated technical roles in tech.
What research scientists actually do
The core of the work is research output — papers, code releases, methodological contributions — produced with enough rigour to advance the field:
Original research agendas. Most senior research scientists own one or more research themes: a direction they're pushing over a multi-year horizon. The agenda has to balance scientific novelty, tractability within current compute, and relevance to the lab's strategic goals. Owning a credible agenda is the single biggest differentiator at senior levels.
Experimental design and execution. The day-to-day work is running experiments: training runs, ablations, scaling studies, evaluations. Experimental craft matters enormously. Strong research scientists design experiments that isolate what they're actually testing; weak ones produce impressive-looking results driven by confounds.
Paper writing and publication. Top-venue publication (NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR) is still the currency in most ML research. Writing takes more time than people outside the field realise — a year of experimental work often maps to six weeks of writing, revision, and rebuttal.
Open-source and artefact release. Increasingly, the artefact (a model release, a dataset, a library) matters more than the paper. Research scientists at labs like Meta FAIR, Anthropic, and EleutherAI spend substantial time preparing releases that will actually be adopted.
Peer review and community participation. The conference review cycle is a significant time tax. Strong research scientists review papers rigorously, attend conferences, speak on panels, and engage with the broader research community. The role is social in ways pure IC engineering roles aren't.
Mentoring junior researchers and PhDs. Senior research scientists typically supervise junior researchers, interns, and sometimes external PhD students. The mentoring load compounds as seniority increases.
How remote research science actually works
The work is heavy on writing, experimentation, and long-cycle thinking — all async-friendly. Compute access through cloud or internal clusters makes the location of the human working on it essentially irrelevant for most projects. Anthropic, DeepMind, EleutherAI, and most non-frontier research groups have material remote hiring; the frontier-most labs still prefer hybrid for specific teams working on capability research with strict security expectations.
The real remote challenge is informal research conversation density. Research environments benefit enormously from hallway discussion, whiteboard debate, and fast iteration on half-formed ideas. Remote teams mitigate with dedicated discussion slots, pair-think time, and heavy use of shared whiteboards (Excalidraw, Miro, Figma), but it is the single most cited complaint from remote research scientists.
The four employer types shape the job
Frontier AI labs — capability research. Anthropic, OpenAI, Google DeepMind, Meta FAIR, Mistral. The work is on the actual frontier of model capability. Teams are small, impact is high, compensation is at the top of the market. Long research cycles; substantial internal debate about direction.
Frontier AI labs — safety / alignment research. Same labs, distinct teams. The charter is to identify and mitigate risks from advanced AI systems. Different publication norms (some work stays internal), different evaluation approaches. Mission-driven culture; substantial philosophical engagement.
Industrial research labs. Microsoft Research, IBM Research, Nvidia Research, Apple ML Research, Meta AI, Google Research. Larger organisations with more formal structure. Broader research scope beyond pure capability. Career ladder is clear. Publication expectations are high.
Applied research groups at product companies. Netflix, Spotify, Airbnb, TikTok, Stripe. Research teams embedded in product-driven companies. Research agenda is typically anchored to business problems. Publication is supported but not central. Strong career homes for researchers who want impact over prestige.
What separates strong candidates
A defensible research agenda. Not "I'm interested in [area]" but a crisp position on what problem matters, why it's tractable now, and what a 12-month research plan against it would look like. The hiring loops at serious labs filter heavily on this.
Technical craft at scale. Running 8-GPU experiments in a notebook is not the same as running 10,000-GPU training runs with coherent scaling analysis. Candidates who've actually operated at scale bring insights that cannot be simulated from smaller-scale work.
Honesty about negative results. Research is mostly things that didn't work. Candidates who can describe their strongest negative result and what they learned reveal more than ones who only recount wins.
Collaboration instinct. Lone-wolf research is increasingly rare. The biggest results in ML are now produced by teams; candidates who can describe how they've led or been part of successful team research — and how they share credit — become senior-leader candidates.
Writing quality. Papers, blog posts, internal memos — the research scientist's output is substantially text. Candidates who write with unusual clarity compound influence across the field. This skill is undervalued in hiring rubrics relative to its actual importance.
Pay and level expectations
US total compensation: Research Scientist I (PhD fresh / 0–2 yrs post-PhD): $220K–$340K. Research Scientist II (2–5 yrs post-PhD): $320K–$500K. Senior Research Scientist (5–10 yrs): $450K–$700K. Principal / Staff Research Scientist: $650K–$1.2M. Frontier-lab compensation at senior levels regularly includes substantial equity grants that can exceed the base+bonus numbers.
Europe adjustment: 20–30% lower base. Top-tier AI labs hiring into London, Paris, and Zurich often pay much closer to US numbers, particularly at senior levels.
Domain premium: Alignment and safety research compensation has pulled roughly even with capability research compensation as the field's strategic importance has grown. Application-oriented research roles (recommendation systems, search quality) pay below pure ML-research roles at equivalent levels.
What the hiring process usually looks like
Typical sequence: recruiter screen, hiring manager call, research interview (discuss a published or in-progress paper in detail), a technical depth interview on ML fundamentals, a research proposal round (what would you work on here and why), panel with senior researchers, final with lab leadership. The research proposal round is the decisive signal — it reveals originality, tractability instinct, and research judgement all at once.
Red flags and green flags
Red flags — slow down:
- No clear research charter for the team. You'll drift.
- Publication is nominally supported but the last paper from the group was two years ago.
- Compute budget is ambiguous or gated by many layers of approval.
- Senior researchers in the group have turned over rapidly in the last 12 months.
Green flags:
- Named research director with credible recent publications.
- Recurring internal research reviews, paper-reading groups, and open tech talks.
- Compute budget is defined and accessible without friction.
- Clear research-engineering partnership model.
Gateway to current listings
RemNavi aggregates remote research scientist jobs from company career pages, AI-lab hiring portals, and specialised ML job boards. Each listing links straight through to the employer to apply.
Frequently asked questions
Do I need a PhD for research scientist roles? At frontier labs and industrial research labs, almost always — or equivalent demonstrated research output (first-author top-venue publications, significant open-source research contributions). Some product-adjacent research teams hire strong MS+experience candidates. Read the listing; don't assume.
How is research scientist different from applied scientist? Research scientist work is measured primarily by research contributions (papers, novel methods). Applied scientist work is measured primarily by shipped product impact. Frontier labs blur the line; most other organisations draw it cleanly.
Is now a good time to pivot from academic research into industry research? Yes, particularly in ML. Industrial labs have compute access that most universities can't match, compensation advantages that have grown considerably, and research problems of material importance. The trade-off is less freedom than an academic position and sometimes stricter publication review.
How many papers should I have published before applying to frontier labs? There's no hard number, but two to four first-author papers at top venues is typical for new-PhD hires at frontier labs; for mid-career hires, the bar is usually a defensible research track record that senior researchers in the lab already know. Networks matter: citations, conference attendance, and known relationships measurably help.
Can I do research-quality work outside a lab? It's harder but not impossible. The EleutherAI, LAION, and Hugging Face communities have produced legitimate research outside traditional lab structures. Compute access remains the hardest constraint. Joining an organisation that provides it is still the dominant path.
RemNavi pulls listings from company career pages and a handful of remote job boards, then sends you straight to the employer to apply. We don't host the listings ourselves, and we don't stand between you and the hiring team.
Related resources
- Remote Applied Scientist Jobs — Product-impact counterpart to pure research
- Remote ML Engineer Jobs — Production-engineering partner role
- Remote AI Engineer Jobs — Application-layer ML role
- Remote LLM Engineer Jobs — LLM-specialist engineering track
- Remote Data Scientist Jobs — Business-analytical adjacent role