Featured competition · National · Vendor-neutral
The Great IndianFDE HiringHackathon 2026
Ship the agent that cleans up a 43-year-old manufacturer — and prove it by the business outcomes it delivers. Four task families, one shared environment, and a hidden key that decides the score. Download the data, build in whatever environment you already use, and upload the output file for automatic scoring.
- Entrants
- —
- Submissions today
- —
- Hours to next deadline
- —
- Host
- AIM Media House
Opens 20 Aug 2026 · Closes 27 Sep 2026, 23:59 IST · Trophy + AIM goodies · Cypher 2026 stage · Certificates for every scored participant
Build an Agent for the Enterprise That Never Got Rebuilt
You have joined a 43-year-old manufacturing company where technology has accumulated rather than evolved.
The ERP is decades old. The CRM migration was never completed. Thousands of service tickets remain open. Critical business decisions live inside shared email threads. Policies are scattered across more than a thousand documents—and sometimes contradict each other.
Your challenge is to build an Agentic AI solution that can operate inside this environment.
The agent must demonstrate its ability to work across four enterprise challenges:
Connect information across fragmented systems and identify inconsistencies, duplicates and mismatches.
Retrieve the right enterprise knowledge, resolve conflicting information and provide evidence-backed answers.
Execute multi-step workflows across systems while maintaining context, state and business rules.
Recommend or execute changes without breaking downstream processes, while recognising when human approval is required.
Your Mission
Build an agent capable of understanding the environment, reasoning across systems, selecting the right actions and producing the required business outcome.
You may use any models, frameworks, agent architecture or development environment.
The technology stack is your choice.
The outcome is not.
Download the challenge dataset, build your solution and submit the required output file for automated evaluation.
Your solution will be tested against unseen scenarios and a hidden evaluation set designed to measure whether your agent can reason and operate reliably—not simply reproduce expected answers.
The winning agent will not be the one that talks best.
It will be the one that can operate inside the mess.
Registration opens
The client
Four decades of accumulated systems. Nothing was ever fully decommissioned.
Legacy ERP
commissioned 1998, patched ever sincePurchase orders, inventory, and the general ledger. Batch jobs run at 02:00 and nobody knows who wrote them.
Half-migrated CRM
migration started 2022, never finishedAccounts exist in both the old and new instance. Neither is authoritative. Both are queried.
Ticketing queue
3,400 open ticketsCustomer escalations, warranty claims, and dispatch requests, loosely categorised at best.
Email inbox
shared ops@ mailboxApprovals, invoice disputes, and vendor chases live here as free text and forwarded threads.
Document store
1,100+ PDFs and scansThe 2019 pricing policy contradicts the 2023 one. Neither is marked superseded.
The four enterprise challenges
Your agent faces all four inside the same environment, and is judged on the outcome of each.
Reconcile
Connect information across fragmented systems and identify inconsistencies, duplicates and mismatches.
Understand
Retrieve the right enterprise knowledge, resolve conflicting information and provide evidence-backed answers.
Orchestrate
Execute multi-step workflows across systems while maintaining context, state and business rules.
Act safely
Recommend or execute changes without breaking downstream processes, while recognising when human approval is required.
How to compete
The leaderboard measures what your system got right. How you built it is your business.
Evaluation
Automatic re-runs
Every submission is re-run against a hidden task suite. The leaderboard updates itself — no manual judging round.
Correctness first
Correctness and completion dominate the score. Cost and latency only matter as tiebreakers.
Metered proxy
All model calls run through a metered proxy with a per-submission cost cap, so engineering wins over spend.
Any model, any framework
Vendor-neutral by design. Bring whatever stack you would actually deploy at a client.
Rewards
Top three finishers take home AIM goodies. Rank 1 also lifts the Great Indian FDE Hackathon trophy on stage at Cypher 2026.
Winner
- ·The Great Indian FDE Hackathon trophy by AIM
- ·AIM goodies hamper
- ·Winner certificate, handed over on stage at Cypher 2026
First runner-up
- ·AIM goodies hamper
- ·Runner-up certificate
Second runner-up
- ·AIM goodies hamper
- ·Runner-up certificate
Every participant with at least one valid scored submission receives a digital participation certificate from AIM Media House.
Why engineers should compete
Prove it on a real enterprise mess
Not a toy dataset. Your agent is scored by machine on outcomes inside a fragmented 43-year-old estate — no slides, no pitch round.
A stage moment at Cypher 2026
Finalists walk through their agent live and the winner lifts the trophy in front of India's AI industry at KTPO Whitefield.
It is a hiring hackathon
Your profile and results are visible to hiring partners looking for forward deployed engineers, and they can reach out directly.
Certificates and a public rank
Every scored participant gets a certificate, and the live leaderboard is a shareable record of where you placed nationally.
Any model, any framework
Vendor-neutral by design. Bring the stack you would actually deploy at a client — nothing here is locked to one provider.
Timeline
- P020 Aug 2026Registration opens
- P106 Sep 2026Registration closes
- P220 Sep 2026Qualifier submissions close
- P302 Oct 2026Finalists announced
- P407–09 Oct 2026Finals + winners at Cypher 2026