Snowflake · Primly Community

how I'd prep for the Snowflake interview if I started over, honestly

backend_bekah · 4 replies

Went through the Snowflake SWE loop twice. Failed the first time, got the offer the second time. Here's what I'd do differently from day one if I had to start fresh.

Coding prep: They run Leetcode-style rounds. Medium difficulty is the baseline, but they do give harder problems to senior candidates. Graphs, trees, sliding window, union-find. The problems aren't exotic but the time pressure is real and the interviewers are watching how you communicate while you solve, not just whether you solve.

If I started over: I'd grind Neetcode 150, specifically the patterns, not the individual problems. Know why you're doing what you're doing, not just that it works.

System design: This is where I lost the first loop. They care about Snowflake-relevant problems: data warehousing, distributed query execution, ingestion pipelines, columnar storage. I went in with generic distributed systems knowledge and got exposed the moment they asked about query optimization across a massive columnstore.

If I started over: I'd read Snowflake's public architecture papers and blog posts. There are a few engineering blog posts they've published on their virtual warehouse model and how they separate storage from compute. Read those. Know the vocabulary.

Behavioral: They use a competency rubric. Collaboration, ownership, drive. Every behavioral question is trying to evaluate one of those. Think through your strongest stories on each dimension and be ready to give the short version (2 mins) and the long version (4 mins) depending on what the interviewer wants.

My first loop I told interesting stories. My second loop I told relevant stories. The difference is knowing what they're actually evaluating.

The misc stuff that also matters: Research the product. If you interview for a platform team, know what Snowpark is. Know the difference between Snowflake's approach to governance vs. competitors. You might get asked, and even if you don't, it signals you're genuinely interested and not just interviewing everywhere on spray-and-pray.

Total prep time second loop: about 6 weeks, 1-2 hours per day. Felt like the right amount.

4 replies

marketer_mei

The architecture paper tip is underrated. I did something similar when interviewing at Databricks and knowing the actual system design made the SD round feel almost collaborative instead of adversarial.

sec_sasha

What coding environment do they use? HackerRank, Coderpad, something internal?

content_cole

When I went through it was Coderpad. They let you use your own IDE if you prefer but most people just use the Coderpad environment. Run your code, it's fine. Nothing fancy.

pivot_pat

"Interesting stories vs. relevant stories" is the thing. I've fumbled behavioral rounds for exactly that reason. The story is good, just not the right one for what they're measuring.