AI Research Scientist, fully remote, full-time, $165,000 a year, open to candidates anywhere.
Background required
This role generally requires a master's degree in computer science, machine learning, or another quantitative field. A number of candidates who apply also hold a doctorate, though it isn't required, and a strong applied research track record can carry as much weight as an advanced degree on its own. In addition to the education requirement, this role calls for three years of research experience, backed by publications or applied research projects demonstrating real depth rather than surface familiarity.
Strong mathematical fundamentals matter here in a way that's easy to underestimate. Someone who can implement a paper's method in code but can't reason clearly about why it works, or where it's likely to break down, will hit a ceiling fast in this role.
The three years should reflect genuine research depth, not adjacent applied ML work with an occasional experiment attached. A candidate whose experience is mostly building and tuning production models, without much time spent designing original experiments or reading deeply in a specific research area, will find the technical bar here a real stretch rather than a natural fit.
Responsibilities
- Design and run experiments to advance machine learning methods
- Publish findings, whether internally, externally, or both, depending on the project
- Translate research breakthroughs into applied models that actually reach production
An ablation study on one recent project revealed that a component the team had assumed was doing most of the work barely mattered at all. Removing it entirely changed final accuracy by less than half a percent. That single result reshaped the direction of the following quarter's research, since it meant the team had been optimizing the wrong part of the system for months. Being willing to follow that kind of result, even when it undercuts an assumption the whole team had built work around, is core to this job.
You'll also work directly with engineering to prototype and validate novel approaches before they're built into production-ready solutions. That handoff is often where a research idea either proves itself under real constraints or reveals a gap nobody caught in the lab setting.
Not every experiment leads somewhere. A meaningful share of the work here ends in a negative or inconclusive result, and documenting why an approach didn't pan out is treated as real progress rather than a failure to hide. Knowing when to stop pursuing a direction is as valuable a skill as knowing which direction to pursue in the first place.
Skills
- Deep learning
- Python
- PyTorch or TensorFlow
- Statistics
- Research methodology
- Academic publishing
- Experiment design
- Mathematics
Academic publishing experience matters even for candidates whose long-term plans point toward industry rather than academia. Writing up results clearly enough for peer review forces a level of rigor that's hard to fake, and that discipline shows up directly in how well someone structures experiments here.
Pay and benefits
This role pays $165,000 a year. Coverage includes comprehensive health insurance, paid time off, and retirement plan contributions, as well as dedicated support for conference attendance and publication costs. Access to significant compute resources is also part of the package, since a lot of the experimentation this role involves would be impractical to run on standard team infrastructure.
Naukri Mitra is coordinating hiring for this opening, and compute access, in particular, has been used by past hires for training runs that would have taken weeks on more limited hardware.
Reviews happen annually and are based on the significance of what someone's contributed, rather than on a fixed schedule of raises. A researcher whose work directly shaped a shipped product feature, or who published something that shifted how the team thinks about a problem, tends to see that reflected clearly in the next cycle.
Anyone comparing AI research scientist remote salary figures at the three-year experience mark should find this offer well above the median, reflecting both the seniority of the role and how competitive hiring has become for researchers who can move fluidly between publishing and shipping.
How the research fits into the business
This isn't a purely academic research position. Findings here are expected to eventually connect back to something the company ships, even when the immediate work is exploratory. That expectation shapes project selection more than it might at a pure research lab, and candidates who want research entirely disconnected from applied outcomes may find this role a different fit than expected.
Among AI research scientist jobs worldwide, the split between pure research and applied research varies enormously by employer. This one sits closer to the applied end, with roughly a quarter of the time built in for genuinely open-ended exploration, while the rest stays anchored to problems the broader engineering org actually needs solved.
Publication isn't treated as a vanity metric here, but it's not mandatory for every project either. Some work gets written up for external conferences. Other work stays internal because it's tied too closely to product specifics to be worth publishing, and that decision gets made project by project rather than following a fixed quota.
Applying
Send a resume along with a link to a paper, project writeup, or applied research result you're proud of, and a short note on what you'd want to work on here. Interviews include a technical presentation of past research, a deeper discussion of experimental methodology, and a conversation about applying research to production constraints.
