SHI International Interview Questions
& Process

It was a really casual chat, and I got a much clearer picture of the job and how the team is set up. The person interviewing me was super upfront and took their time to explain the job in detail, plus answered my questions really well. Even though the talk was helpful and gave me info, I realized the role wasn't really what I was looking for based on my own interests and where I want to go. All in all, it was a good experience, and the interviewer was professional, open, and easy to talk to.

It starts with a HR round, then the hiring manager, a team member, an assignment, and finally a round with leadership. After all the rounds, HR sent a thank you note and mentioned they would continue searching for other candidates. I think it's a waste of candidate time and energy. SHI also gets assignments done for free, which could help their future strategies. My suggestion: have a clear discussion about what they are looking for in a candidate and only proceed with the assignment round if there's a match. It's exhausting, and SHI ends up taking candidates for a ride.

I had a disappointing experience with this interview process. While the interviewers were polite, the entire thing felt disorganized and not transparent. Communication was slow, and there were big delays between steps without any clear updates or expectations. During the interviews, some questions seemed unclear and repetitive, and the interviewers didn't give much context about the role or how the team is structured. It felt like they weren't aligned on the skills they were actually looking for. At times, the discussion felt rushed, and I didn't get to fully explain my experience. The recruiter communication also fell short - responses were delayed, and follow-up information was inconsistent. Overall, the process lacked structure and clarity, which made the whole experience frustrating. Pros: Interviewers were polite. Cons: Very slow communication, no clear explanation of role expectations, unstructured interview format, repetitive or unclear questions, lack of transparency throughout the process. Final Thoughts: This interview process could be improved with better communication, clearer expectations, and more structure. My experience did not reflect the professionalism I expected from a company of this size.

The interview process started with a recruiter call followed by a hiring manager screen where I was asked about ML, software, and behavioral topics, which seemed appropriate for the role. Things got complicated in the next round with a compute-intensive, writing-heavy, and time-consuming NLP project assigned as a take-home task with a one-week deadline, supposedly to assess 'problem solving skills.' After they reviewed the code, I was invited for an onsite interview. Unfortunately, I was rejected after the onsite, with the feedback being 'bugs in the code.' It's baffling why the code ran on my end but had bugs on theirs, or why they'd proceed to an onsite if code bugs were a disqualifier, especially a week-long project. They asked for a production-ready, bug-free project in a week, assuming candidates have supercomputers and can take time off work. They also expect candidates to be experts in all ML applications, not just specific areas like LLMs, calling the role 'Senior Machine Learning Engineer' but focusing on LLM specifics. During the onsite, I had to run a compute-intensive preprocessing and evaluation pipeline using libraries like Spacy and BERT on an 8GB MacBook Air within 30 minutes, which seems unreasonable. They could have checked logs or discussed any issues during the onsite, especially since the code worked fine on my machine. The expected solution was to use a pre-trained model from Hugging Face, which feels like a shallow definition of problem-solving for an experienced ML engineer. The need for a detailed specification document is also questionable if the focus is solely on using Hugging Face. A simple Q&A about using ChatGPT or zero-shot methods might have sufficed. It's unclear why they'd provide packaging, sharing code, documentation, and logs if they wouldn't review them thoroughly, especially when expecting candidates to dedicate a full week. Any bugs arising from transitioning code from Jupyter notebooks to modular Python scripts should have been addressed during or before the onsite.