The Portfolio Project That Gets You Hired

Hey, both interviews went well 🙏 beON had quite a lot of questions on ML system design so I answered them using my project structure. The team lead also said he had a look at my project earlier! Enercity: showed my deployed product and since the team was working on grid use cases, they had quite a lot of questions. Next week I’ll hear from both of them. I’ll keep you posted :)

Aleenaaa just got the second offer 🥳 The hiring manager pulled up their pipeline diagram and asked how I’d integrate the Bosch CNC project into their setup. I literally walked him through the README from Day 3. He said “this is exactly what we’re trying to build internally.” Negotiating between both this week. Will share the salary stuff once one is signed 🤝

Final round done — fingers crossed 🤞 The technical lead spent almost 30 mins on the cancer-detection repo. Barely 5 mins on the model itself — the rest was MLOps: how I set up DVC, why the experiment-tracking is structured that way, how I’d catch data drift in production. I was nervous about that part but Day 3 prep saved me. Should hear by next Friday.

Third interview booked this week 😅 didn’t expect that pace. The recruiter from the Heidelberg biotech literally said “we don’t get many candidates from a research background who can also explain MLOps to us.” I think I underestimated how rare the combo is. The README + the confidence-scoring writeup is doing a lot of work in screening. They’re treating it as 80% of the interview prep.

First-round done ✓ The live demo was the thing — when I dropped the link in the application they passed it around the team. The lead engineer asked how I’d handle device drift if the firmware changed mid-deployment. We talked through it for 20 minutes. He told me afterwards that the deployment alone put me ahead of the rest of the shortlist. Second round next Tuesday. Trying not to overthink it.

Second round done with the Frankfurt fintech 🤞 They were genuinely impressed that the pipeline runs on real Bundesbank macro indicators, not synthetic data. The senior data scientist asked how I’d extend it to predict cashflow at a client level — I sketched out the approach on the whiteboard and he said “that’s basically what we’re trying to build next quarter.” Final round next week. Holding my breath.

First PR review at the new role just came in 😅 Two comments. Two. Used to be 30+ when my “code” was one long script with hardcoded paths and zero tests. What the sprint actually taught me wasn’t modelling — it was professional-grade code structure. Modular code, configs, unit tests. That’s the thing reviewers care about. Different planet from where my GitHub used to be.

First time someone said “show me a project you’re proud of” and I didn’t freeze 🥹 Pulled up the live demo, walked the interviewer through it, ran an actual prediction. He leaned in. Used to be I’d mumble something about “notebooks somewhere” and watch the energy leave the room. Now I’m demoing a live system. Completely different conversation. Got the offer two days later.

Walked out of an interview thinking I bombed the model question 🫠 But the conversation never went back to it — the interviewer kept asking about the deliverables, the pipeline, how the project would slot into a team’s workflow. Old me would’ve led with accuracy numbers. New me led with “here’s how this would actually be useful to your team.” That’s the shift the sprint actually teaches you.