Traditional screening/shortlisting methods and tools were built for a world where quiz scores predicted job performance.

That world no longer exists.

Today's developers and engineers ship with AI copilots, debug production systems, make real-time architectural tradeoffs, etc. — none of which show up in 30-question algorithmic coding challenges or take-homes.

Utkrusht was built from the ground up around one core belief: the only accurate way to evaluate tech candidates is to actually watch HOW they solve a real problem in a real environment, live.

In other words, simply watch-them-work.

1 task is an actual on-the-job task where candidates solve, debug, design, extend, refactor live apps, and deploy it in a live production environment.

We fundamentally believe in the age of AI, watching someone live solve real on-the-job production problems is the best way to accurately identify how good they really are.

In this process, allow them all the tools (including AI) of their choice and then become a fly on the wall!

for eg: For a -

• Fullstack role, candidates are asked to debug and fix a broken API

• SRE role, candidates are asked to write a runbook for incident response

• AI Engineer role, candidates are asked to improve embeddings in a chatbot

• Devops role, candidates are given a broken K8 cluster to fix, optimize a docker container, etc.

• Data Engineer role, candidates are asked to fix kafka partitioning, optimize a hadoop cluster, etc.

• Backend role, candidates are asked to live migrate a DB schema, refactor a payment microservice, etc.

• and many more…

  
  

These tasks also allow you watch candidates how they debug, what decisions they made, what approach they follow, how they break down ambiguous problems, make tradeoffs, etc.

Tasks are designed to measure in-depth fundamental skills, use of AI, and problem-solving approach, making them actually build something live in-action and you watching their recorded session.