Five rounds at DeepInfra, and a reason

Company	DeepInfra (Series B, AI inference)
Team	Software Engineering, Inference
Role	SWE, experienced
Location	Rounds 1 to 3 remote, 4 and 5 onsite in Palo Alto, CA
Date	April 2026
How it started	Cold application
Rounds	5
Outcome	Rejected, with feedback

The other side of the Amazon loop. Same result, completely different experience of it.

The loop

The first three rounds were remote. The last two were onsite in Palo Alto.

Round	What it was	Problem	Self Rating
1	Intro call with the recruiter	The role and the team
2	Problem solving	Graph, cycle detection. Medium to hard	Hire
3	Problem solving	Matrix, combinatorics and graph. Medium to hard	Hire
4	Fit, with the recruiting lead	Experience and background	Hire
5	Problem solving, with a cofounder	Place tiles on a plane. Hard. Plus questions on AI inference	Lean Hire

Worth knowing who was on the other side of the table: nearly everyone on that team is an ICPC or IOI medallist. That explains the shape of the problems. Cycle detection, combinatorics on a matrix, and placing tiles on a plane are not standard interview questions, they are olympiad questions, and they were being asked by people who had won at that level.

Round 5 was the one

At a company this size a cofounder’s read carries more weight than anyone else’s, and I did not solve the tile problem correctly. My communication was good throughout, which is not the same thing as arriving at the answer.

It is tempting to file this under bad luck, a problem arriving from an angle I had not seen. Some of that is true. But writing it down next to who was asking, the more honest version is that the problem was pitched at a level the room had already cleared and I had not.

What they actually said

The stated reason was that they were looking for someone with more experience for this role.

What made this good anyway

I was told why. That sounds like a low bar and it is, but it is one that a great many companies do not clear. Four rounds of real technical conversation ended with a specific reason instead of silence, and the reason was actionable.

I am still in touch with most of the people who interviewed me, including the founding team and the recruiter. That is the part that matters more than the outcome. A loop that ends with a rejection and five new professional relationships is not a loop I regret spending a month on.

I flew out for this one too. Compare it with flying to Seattle for four rounds and learning I had been rejected by inference. Two companies, the same amount of my time and travel, the same answer at the end. One of them told me why.