A complete waste of time. Before you even speak to a recruiter, you're expected to spend four days on a pre-assessment where you deploy a model using an API framework. If you pass, you move on to a 1-hour panel with senior MLEs, supposedly to discuss your work. The final round includes three more 1-hour interviews: technical (likely system design), behavioral, and open QA.
I didn’t make it past the second round. The recruiter emphasized that it would be focused entirely on the pre-assessment—to evaluate ownership, decision-making, and depth of understanding. Makes sense, considering the time commitment. But the MLEs clearly hadn’t looked at the project. The interview ended up being more a blend of what I would expect during the final round (technical/behavioral).
Don't waste your time on the assessment, here are some repos that show the kind of project they expect, beedabug-mle, OMS1996-state-farm, JJD129-SF-MLE
Domande di colloquio [1]
Domanda 1
They asked a ton of multi-barrelled questions to fully encapsulate everything instead of having a back-and-forth conversation.
Ho presentato la mia candidatura online. Ho sostenuto un colloquio presso State Farm
Colloquio
After the initial interview, I invested significant time and energy into completing the take-home project. We I did not receive any feedback feels like my efforts were undervalued. Requiring candidates to complete a substantial project without providing feedback is a waste of both the candidate's and company's time.
Domande di colloquio [1]
Domanda 1
classification coding exercise that need to be completed in 3 days
Ho presentato la mia candidatura online. La procedura ha richiesto una settimana. Ho sostenuto un colloquio presso State Farm (New York, NY) nel mese di giu 2022
Colloquio
Submitted a resume, they responded a couple of days later with a take-home.
The assignment was ludicrous and convoluted. The code included was full of really horrible (as in, start spitting out bogus data in prod) bugs, unused variables, poor structure, etc. The goals/outcomes were spread across multiple files that contradicted themselves.
In the end I was rejected, and they refused to give any feedback. If you give a take home, at least have the decency to review it with the candidate.
I feel like I dodged a bullet.
Domande di colloquio [1]
Domanda 1
Some really poorly written Jupiter notebook. Deploy it and make it scalable.
