SoundHound Machine Learning Engineer Interview Questions
& Process

The recruiting process was clear, but it took many rounds over several months. There were a lot of coding interviews and a few more about machine learning and statistics. The interviewers were super friendly, but demanding and challenging. It was a nice experience overall, but a very long one, which was balanced by the fact that the interviewers made me want to work there.

The interview process had 4 stages. The first stage was an online software knowledge assessment, which was tough and covered almost all computer science topics with multiple-choice and coding questions. This was followed by an HR phone screen where we discussed the company's product, vision, and my job responsibilities. The HR person was very professional and nice. Next, there was a video interview with the VP, which included 3-4 coding questions and 1 probability question, done using coderpad. The final stage was an onsite interview at the Santa Clara HQ, consisting of 5 rounds, each about an hour long, with a lunch break with the team in between. The whole process was smooth and efficient, with professional and nice interviewers and HR. It was a very enjoyable experience.

The application experience was negative, but the role and mission were exciting. The company's mission in Voice AI is genuinely exciting, and the role itself (owning the end-to-end TTS stack) is highly appealing. However, the initial pre-assessment phase is poorly designed and misleadingly described. The recruiter email stated the assessment would test "hands-on coding proficiency." In reality, the test was highly inconsistent and covered a vast, unfocused range of topics. It was a mixture of obscure/basic theory (e.g., specific definitions in linear algebra or obscure ML terms) that are irrelevant for a Senior Engineer focused on model design and production deployment, tricky/unclear questions that were ambiguously worded, overly complex, or designed to trick the candidate rather than test deep, practical understanding, and advanced Python coding that was indeed challenging and appropriate for a senior role, but its effectiveness was diminished by the unnecessary theoretical noise. This process fails to accurately screen senior candidates and penalizes those who focus their preparation on modern, practical, production-oriented skills (which the job description claims to require). The assessment needs to be completely re-scoped to reflect the actual demands of a Senior ML position at SoundHound.