WWT Unravel 2026: a solo Top 10, and an interview that read my resume back to me

WWT Unravel is World Wide Technology's national campus challenge. I entered solo, which saves a great deal of time on team meetings, and finished in the Top 10. Finalists earned a pre-placement interview. Here is how each round went, and the lesson I paid for in the last one.

• finish

  Top 10

• entry

  Solo

• ML round F1

  ~75%

• OAs cleared

  2

The first online assessment paired multiple-choice questions with data structures and algorithms problems. Years of arguing with Codeforces (Specialist) and CodeChef (3-star) finally paid rent here, and I cleared every problem.

The second round was hands-on ML: train an XGBoost model on financial data to estimate how likely a business was to fail. I landed around 75% F1.

F1, not accuracy, is the right lens here. Failing businesses are the rarer and more important class, and a model can look wonderfully accurate by cheerfully predicting that everyone survives. Optimism is a lovely personality trait and a terrible classifier. F1 only rewards the model for actually finding the failures without crying wolf.

The pre-placement interview covered four areas: the projects on my resume, my forecasting work from the Adani hackathon, classic ML algorithms, and logic puzzles. The ML theory and the puzzles went well, and I got them right.

The resume deep dive did not. The interviewer picked lines off my resume and asked me to walk through exactly how each thing worked, then asked why, then asked why again. Some lines described work I had done alongside teammates, or results I could state but not fully unpack. Those were the moments I wobbled, and I did not convert the interview. Turns out a resume is not a highlight reel. It is a list of things you have volunteered to be cross-examined on.

• Every line on my resume now passes one test: can I explain it on a whiteboard, trade-offs included, without notes? If not, it goes, or it shrinks to the part I genuinely own.
• Team work is described as team work. A crisp "I built X, my teammate built Y" earns far more respect than a vague "we built everything".
• Every project gets a post-mortem. The Adani post on this blog came out of exactly this exercise, and it turned up a validation leak I now catch on day one.
• I rehearse the follow-ups, not the pitch. Why this model? What else did you try? How do you know it worked? What broke?

The Top 10 proved I could do the work under pressure. The interview proved that explaining the work is half the job. I've since rebuilt my resume and my public repos around that idea, and this blog is, in a sense, the receipts.