Lucio_Challenge

We’re exploring how far large-scale document reasoning can really go under real constraints. The Lucio Challenge is where we push those ideas to their limits.

Total Prize Pool : $7000

15 March

Lucio HQ

10:00 AM – 8:00 PM

Merch for shortlisted candidates

// Lucio Challenge Invite
const Challenge = {
  name: "Lucio_Challenge",
  date: "March_15_2026",
  mode: "In-Person",
  location: "Lucio_HQ",
  prizes: [
    {
      title: "Evaluate under 30 seconds",
      prize: "$5,000",
    },
    {
      title: "Fastest evaluation",
      prize: "$1,000",
    },
    {
      title: "Most innovative architecture",
      prize: "$1,000",
    },
    {
      title: "Default",
      prize: "Fun merch and swag",
    },
  ],
  whoCanJoin: [ "developers", "designers", "students", "everyone-who-builds"],
  perks: ["mentors", "workshops", "swag", "networking"],
};

// Lucio Challenge Invite
const Challenge = {
  name: "Lucio_Challenge",
  date: "March_15_2026",
  mode: "In-Person",
  location: "Lucio_HQ",
  prizes: [
    {
      title: "Evaluate under 30 seconds",
      prize: "$5,000",
    },
    {
      title: "Fastest evaluation",
      prize: "$1,000",
    },
    {
      title: "Most innovative architecture",
      prize: "$1,000",
    },
    {
      title: "Default",
      prize: "Fun merch and swag",
    },
  ],
  whoCanJoin: [ "developers", "designers", 
"students", "everyone-who-builds"],
  perks: ["mentors", "workshops", 
"swag", "networking"],
};

Main

Lucio_Challenge_2026

One_problem_worth_solving.

One_problem
_worth_solving.

Many_minds_at_work.

Many_minds
_at_work.

Many_minds_
at_work.

{What_Is_This_Challenge?}

{What_Is_This_
Challenge?}

200 Document Corpus. 15 Questions. 30 Seconds.

Build the fastest document ingestion pipeline

Ingest and structure a fixed corpus of 200 documents for efficient downstream

reasoning.

Answer complex questions against ground truth

Accurately answer 15 predefined questions, evaluated against known correct answers from the full corpus.

Meet strict performance

targets

Deliver end-to-end results in under 30

seconds, balancing latency, accuracy, and robustness.

{What’s_At_Stake?}

{What’s_At
_Stake?}

Talk is cheap. Benchmarks aren’t.

If you can do this in under 30s, you win

$5000

The fastest solution wins

$1000

The most innovative architecture wins

$1000

{What_You_Need_To_Enter}

{What_You_Need
_To_Enter}

Build a prototype against the test corpus and document the strategy behind your approach.

Here is the test corpus with 17 documents that mirrors the final dataset of 200 in structure and complexity,
along with ground-truth answers. Use it to design, test & optimise.   

Build a prototype against the test corpus and document the strategy behind your approach.

Here is the test corpus with 17 documents that mirrors the final dataset of 200 in structure and complexity, along with ground-truth answers. Use it to design, test & optimise.   

{Corpus}

Use this corpus to understand:

The document types you need to ingest
The formats you’ll encounter
The scale and structure of the data

What’s guaranteed:

Each document will be under 50 MB
Only formats present in the test corpus will appear in the final dataset

{Answers}

Ground-truth data is provided to evaluate correctness.

Your system will be evaluated on whether it presents accurate answers, grounded in the source documents.

Questions

Answers

Document Name & Page Numbers of the final answer

{Challenge_Constraints}

Use whatever infrastructure works best for you. You can run your solution on your own machine, lambdas, or wherever else you want.

You’re welcome to use any third-party tools or services, provided you have the necessary licenses/authorisations/permissions to use them.

Just note that you’ll cover the cost of running your solution.

Your solution will need to connect to a publicly hosted Lucio server to fetch the final corpus and submit your answers.

For any clarifications, reach out to challenge@lucioai.com

{Who_Can_Participate?}

{Who_Can
_Participate?}

Open to all engineers;

undergraduates must be in their

final year.

Open to all engineers;

undergraduates must be in their final year.

Individual participants or teams of up to 3 members

Individual participants

or teams of up to

3 members

Individual participants

or teams of up to

3 members

Strong problem-solving and execution mindset required

{Challenge_Format}

{Challenge
_Format}

Phase 1 : Online Submission (March 1)

We release:

Test dataset

Problem statement

Constraints & evaluation criteria

Teams build and submit solutions via our website

Phase 2 : Shortlisting (March 3)

Our team reviews submissions

Selected teams are invited to our office

Phase 3: On-site Sprint & Presentations (March 15)

Final sprint & refinement

Presentations + judging

Winner announcement

{What_You’ll_Submit}

{What_You’ll
_Submit}

Short presentation covering:

System architecture

Time taken to execute

Submit here

{Important_Dates}

6 Feb

Registrations Open

1 Mar

Submissions Due

3 Mar

Shortlist Announcement

15 Mar

Challenge Day

{Why_Participate?}

Work on a real, non-trivial problem

Push the boundaries of high-stakes document intelligence

Win exciting prizes

{Consent_&_Code_of_Conduct}

{Consent_&_Code_of
_Conduct}

By registering, participants consent to photography and videography during the event.

All participants must adhere to a code of conduct ensuring respect, fairness, and collaboration.

By participating, you agree that your solution will be published in the public domain to support open learning and community research.

{Powered_By}

Partner

Jumpstart better data engineering and AI future.

{Lucio_Research_Commons}

{Lucio_Research
_Commons}

Our mission is to advance AI engineering through public research and open collaboration. All solutions will be made public for the community to build on.

Are you up for it?