Are applied MLE skills are not useful at a startup?

Question

Hi All,

I work on ML in Google Ads. I spend most of my time training models and running experiments. I spent little time writing code but more on analyzing the experiments, followed by hypothesizing and designing new experiments.

Most startups are using LLMs from frontier labs or open-source models and prompting them to work for their use case. It feels that my knowledge in building models is not particularly useful because these startups are not building LLMs from scratch, and most of the work is done in prompting, evals, and building infrastructure to support these models. I feel this mainly involves backend skills and the required ML skills like prompting, evals can easily be picked up by general SWEs too.

I feel that, unless you work on frontier AI research, the skills of an applied MLE are not that useful in building or working at a startup.

Would love to hear what people think about this.

Thanks!

Thoughtful Tarodactyl · Accepted Answer

There are plenty of startups that hire for research engineering roles (not frontier LLM research). I interviewed at a few as well. I wouldn't count yourself out.

Right now there are a TON of startups trying to take a slice of the LLM wave so finding good startups with LLM work is just noisy now. I also think fine tuning skills are important because many companies are building fine-tuned LLMs on proprietary datasets (eg medical/financial models)

https://fal.ai/careers/applied-ml-engineer

https://elevenlabs.io/careers/3d650946-5ac2-4729-9ae4-129c43fcd0b5/research-engineer?ashby_jid=3d650946-5ac2-4729-9ae4-129c43fcd0b5

https://jobs.ashbyhq.com/twelve-labs/c3c6c6c2-795c-4eff-a236-5f6059f6d112

Are applied MLE skills are not useful at a startup?

Discussion

Other Great Discussions