Skip to content

In short

Machine Learning is required most often, in 7 of 11 principal research scientist roles at tech companies whose requirements we read in the 90 days to Nov 21, 2026, then LLMs (6) and AI Agents (4). Among roles that state years, 6 of 6 ask for at least 5 years. We counted 11 roles, each at a different employer.

Machine Learning
7 of 11
Required most often, of the roles we could read
5+ years
6 of 6
Of roles stating years
Remote roles
7 of 11
Of the roles we could read for it

What the postings ask for

Out of the 11 roles whose requirements we could read, of 11 in all. Too few to show shares, so we show counts.

Required
SkillRoles asking
Machine Learning
7 of 11
LLMs
6 of 11
AI Agents
4 of 11
Python
4 of 11
PyTorch
3 of 11
Statistics
3 of 11
AI
2 of 11
Algorithms
2 of 11
Computer Vision
2 of 11
Fine-Tuning
2 of 11

Does your resume show these? Review my resume

Preferred
SkillRoles asking
Open Source
3 of 11
Reinforcement Learning
2 of 11
CI/CD
1 of 11
Conversational AI
1 of 11
Deep Learning
1 of 11
Docker
1 of 11
Experimentation
1 of 11
Fine-Tuning
1 of 11
Generative AI
1 of 11
Git
1 of 11

Experience

Asked most often: 5–7 years, 4 of 6.

Of the 6 roles that state years

Under 5 years
0 of 6
5–7 years
4 of 6
8–9 years
0 of 6
10+ years
2 of 6

Degree

4 of 11 require a degree, and 3 more accept experience instead.

Of the 11 roles we could read for it

Required
4 of 11
Required, or experience instead
3 of 11
Preferred
1 of 11
Not mentioned
3 of 11

Pay and remote work

Of the 11 roles we could read for it

Roles stating pay
7 of 11
Remote roles
7 of 11

Questions about principal research scientist roles

Companies with the most research scientist roles

Have a principal research scientist posting in mind?

Paste the posting and your resume to see each requirement checked. No account needed.

No posting yet? Review your resume instead

Check a principal research scientist posting

How we count: 11 roles in the 90 days to Nov 21, 2026, each role counted once however many boards post it. Methodology

Something wrong here?