What to put on a research scientist resume at tech companies · 122 roles · 90 days to Oct 16, 2026
Updated
In short
Of the 122 research scientist roles at tech companies whose requirements we read in the 90 days to Oct 16, 2026, Machine Learning is required in 73 of 122 (60%), Python in 61 of 122 (50%) and PyTorch in 43 of 122 (35%). Name the ones you have in your latest jobs, next to what you built with them.
- Degree required, no alternative
- 38 of 121 (31%)
- Of the roles we could read for it
The skills these postings ask for most
Counts are out of the 122 roles whose requirements we could read; the 10 asked for most in each list.
| Skill | Roles asking |
|---|---|
| Machine Learning | 73 of 122 (60%) |
| Python | 61 of 122 (50%) |
| PyTorch | 43 of 122 (35%) |
| AI | 27 of 122 (22%) |
| LLMs | 26 of 122 (21%) |
| Reinforcement Learning | 24 of 122 (20%) |
| Experimentation | 22 of 122 (18%) |
| Deep Learning | 21 of 122 (17%) |
| Statistics | 19 of 122 (16%) |
| Computer Vision | 17 of 122 (14%) |
| Skill | Roles asking |
|---|---|
| Machine Learning | 14 of 122 (11%) |
| Open Source | 13 of 122 (11%) |
| Reinforcement Learning | 10 of 122 (8%) |
| Generative AI | 9 of 122 (7%) |
| Model Evaluation | 9 of 122 (7%) |
| AI | 8 of 122 (7%) |
| C++ | 8 of 122 (7%) |
| GPU Computing | 6 of 122 (5%) |
| PyTorch | 6 of 122 (5%) |
| Data Pipelines | 5 of 122 (4%) |
What to put first
Lead with Machine Learning, Python and PyTorch
They are the required skills named most, with their counts in the table above. Put the ones you have in your skills line and in the bullets of your latest job, each next to what you did with it.
Then Open Source and Reinforcement Learning, if you have them
Preferred, not required: Open Source in 13 of 122 (11%) and Reinforcement Learning in 10 of 122 (8%). Add them where they are true; leaving one out costs less than claiming it.
Make 3+ years easy to count
Among roles that state years, 39 of 54 (72%) ask for at least 3 years. Give every job a start and end month and year, so a reader can add them up.
Education can go last
38 of 121 (31%) roles require a degree with no alternative; the rest accept experience instead, prefer one or do not mention it. Lead with your work.
What changes by level
No two neighbouring research scientist levels have 50 or more roles whose requirements we read yet, so levels are not compared. The career path lists each level and its roles.
Bullet patterns
Replace every bracket with your own facts. Keep only numbers you can explain in an interview.
Pattern, not a real resume
Built [system] that [what it does] for [team or users], using Machine Learning; [metric] after [time].
Pattern, not a real resume
Used Python to cut [metric] from [before] to [after] by [change you made].
Pattern, not a real resume
Applied PyTorch to [problem], which [made it cheaper, faster or more reliable] by [amount].
Pattern, not a real resume
Led [project] with [number] people across [teams]; shipped [outcome] in [time].
Gaps to check before you send it
A required skill you use but never name
If you work with Machine Learning and Python but your bullets only say what the product did, a reader looking for the skill will not find it. Name it where you used it.
Dates that do not add up
Most roles that state years ask for at least 3 years. Missing months, overlaps and unexplained gaps make that hard to count.
Bullets without an outcome
A bullet that names a tool but not what changed reads as a list of tasks. Add what moved and by how much, or what shipped.
More on research scientist roles
How we count: 122 roles from 85 employers in the 90 days to Oct 16, 2026, each role counted once however many boards post it. Methodology