Prizes & opportunities
A ₹50,000 cash pool at the Grand Finale, a real conversation with the Sydon.AI engineering team, industry recognition, and a network of five hundred builders. Here is exactly what is on the table, and what is not.
The cash
One pool, presented at the Grand Finale on 10 October 2026 in front of the judging panel.
₹50,000
Total cash prize pool
Grand Finale · 10 October 2026
Beyond the cash
Most people who finish CUBE do not finish for the money. These are the outcomes that keep working after 10 October.
Your work is read by practising engineers at every round, not skimmed at the end. Exceptional participants continue the conversation with the Sydon.AI engineering team after the programme.
That conversation is an outcome of demonstrated ability across the whole programme. It is not a guarantee of an interview, an offer or employment, it is never decided by the final demo alone, and we are not advertising a number of roles or a specific position. What we can commit to is that strong work gets read by the people who would actually work with you.
Every submission is reviewed by engineers who build production systems for a living, and the defence is a live technical conversation with them rather than a pitch to a panel of spectators.
Finalists are recognised publicly by Sydon.AI and CodeQuesters. After the Grand Finale you are free to publish your own submission, so the work becomes something you can point at: a repository, an architecture you can explain, and a defence you have already survived.
Around five hundred selected builders, one hundred finalists, twenty pods of five. Round 3 puts you in a pod with four specialists you did not choose, working on real cross-agent dependencies, which is how professional engineering actually feels.
The people you meet include the engineers who ship intelligent commerce for a living. That network does not close when the Finale ends.
CUBE is scored on evidence at every stage: applications on your repositories and commits, Round 2 on working code merged into the starter repo, the Finale on a system that runs live and an individual who can defend it.
What you leave with is a defensible record rather than a participation badge: the “Earn” letter of the CUBE, in the programme’s own terms.
Whether or not you win
The prize pool goes to a few people. This part does not.
Reasoning loops, tool use, grounding and evaluation: the parts that decide whether an agent is useful or merely fluent.
Version control, code review, testing, and working habits that hold up outside a hackathon weekend.
How an independent capability becomes a dependable part of a larger product you do not fully control.
Five specialists, one system, real cross-agent dependencies, and the negotiation that comes with them.
Failure modes, edge cases, security and observable behaviour under real conditions.
Explaining architecture, decisions and trade-offs to people who will challenge them.
Certification
Entry-level verified credential for shortlisted candidates.
Functional prototype submission passing automated CI/CD checks.
Selected among 100 finalists from 1,000+ applicants.
Built and integrated multi-agent AI systems through a production-oriented engineering workflow.
How it is decided
Four points where the panel forms a view. Every one of them is about work that runs.
Applications are read for evidence: your repositories, your commits, projects you can explain end to end. Roughly 500 selected from around 1,000.
One participant, one agent, one domain, inside a starter repository that already runs. Submissions close 27 September: working code, merged. Assessed within each domain, so every pod gets one specialist per challenge.
The 100 finalists become 20 pods of five. Every agent must consume something another agent produces. The pod ships end-to-end workflows no single agent could complete alone.
The system runs in front of the judges, and then you defend it: architecture, decisions, failures, trade-offs, performance, and exactly what you contributed. Judges score the system and the individual. Prizes are awarded here.
The rounds in full: Explore CUBE.
Straight answers