Perplexity

Software Engineer

The Perplexity onsite took place in the San Francisco office and included a coding round, lunch, and a system design interview. If this stage was passed, there was expected to be another project deep dive afterward.

Nvidia

Software Engineer

The NVIDIA GPU Performance team scheduled three technical phone screens from the beginning. The interviews were highly specialized around performance engineering, GPU networking benchmarks, and low-level C memory management rather than standard algorithm questions. The three rounds covered performance engineering and benchmarking concepts, designing a GPU RDMA ping-pong microbenchmark in Python, and implementing aligned memory allocation in C.

Glean

Software Engineer

The first-round Glean TLM interview was a graph problem built around a train schedule. The main challenge was not BFS itself, but modeling the schedule correctly and incorporating time into the traversal.

TSMC

Computer Integrated Manufacturing Engineer

The first-round TSMC Arizona CIM Engineer interview was a roughly 45-minute Microsoft Teams panel with several CIM managers and technical leaders. There were no coding or algorithm questions. The interview focused almost entirely on behavioral questions, detailed resume and project deep dives, motivation for joining TSMC, understanding of the semiconductor industry, and willingness to work in a manufacturing-oriented IT environment. The strongest preparation would be knowing every project on the resume in detail, being able to quantify impact, and having a convincing explanation for why TSMC and semiconductor manufacturing make sense as the next career step.

Cresta

Software Engineer

The Cresta AI-assisted coding round asked me to build a conversational flight search and booking agent using an existing Python starter project, API documentation, and an LLM. The task was very implementation-oriented. I had to set up and run the project locally in my own IDE, use the supplied API credentials, define the appropriate LLM tools, connect those tools to the provided flight APIs, and maintain the state required to complete a multi-step booking flow.

Snowflake

Software Engineer

The Snowflake interview process included two technical phone rounds followed by an onsite. The phone interviews covered system design and coding, while the onsite included an AI-enabled coding exercise and a project deep dive with behavioral questions. The technical topics were fairly practical: designing a data ingestion pipeline from S3 into a database, implementing a transactional key-value store, and building a rate limiter with AI assistance.