Teradata Senior AI Engineer Interview Experience

YOE: 4.5 Recruiter reached out on LinkedIn Round 1: DSA (45 min) Extremely Easy 1. Circular House Robber Made a modification to also find the indexes of the robbed houses. 2. 2D grid, min path sum

YOE: 4.5 Recruiter reached out on LinkedIn Round 1: DSA (45 min) Extremely Easy 1. Circular House Robber Made a modification to also find the indexes of the robbed houses. 2. 2D grid, min path sum from top-left to bottom-right cell. Can only go down or right. Solved both of them ease with 10 min to spare Round 2: System Design (45 min) Weird Started with questions like release cycles and merging of git branches in my current organization. Spent around 10-15 minutes on such questions. Asked a question about ingesting sensor data from multiple devices. The sensor may be located in remote locations. The system should then be able to report aggregated data. There were 3 windows for aggregation: 1 min, 1 hr, 1 day. I suggested the flow md Device -> API Gateway -> Ingestion Service -> Queue Queue -> Consumer -> Postgres - Scheduled jobs querying raw data from Postgres (An index on timestamp would make it efficient). - The jobs would then write aggregated data to their respective Postgres tables. - The raw data can then be deleted if not required anymore. The interviewer did not let me describe my complete solution. He was stuck on why not Device -> Queue directly ? I was like API GW -> Ingestion Service allows for auth, validation and decoupling. He also objected to my choice of Postgres. He suggested me not to use "trendy" technology when any document database could have been used. He lost interest in my solution. Then started asking some AWS related questions. Answered them satisfactorily. Received rejection mail a few hours later. Genuine Questions - Is it common for system to expose their queue directly over the internet ? - Also, a lot a system design sources suggest that Postgres is a good default. Is that no longer true ?