Skip to main content

Meet the winners of the Provn National AI Competition

Builders from across the country took on real challenges for roles AI has transformed. They shipped the work, then recorded a walkthrough explaining how they thought, where AI helped, and where they took the wheel. A panel of AI leaders reviewed every submission and picked the best.

Performance over pedigree. Proof over polish.

winners
9
role ladders
3
scored judge reviews
96

AI Product Manager / Analytics

The challenge: Diagnose an AI feature that is losing customer trust. Set a launch-readiness bar for an AI agent. Make the ship-or-hold call and defend it to leadership.

Headshot of Naman Sharma

1st Place$1,000

Naman Sharma

Four years automating payments operations at JPMorganChase, now a PGDM candidate at SPJIMR aiming for AI PM. Earned the best average rank in the ladder, with three judges placing him first.

What the judges said

“Clear, committed decisions in every wave, each paired with a shippable alternative rather than a binary go or no-go. The Wave 3 readiness bar is the most rigorous in the group.”
Tom Leung · Judge
“Reasoning is sound and decisions and designs are justified. Thoughtful about regressions.”
Neha Monga · Judge
View Naman’s Provn profile
Headshot of KyJahn Smith

2nd Place$250

KyJahn Smith

Software engineer who does paid AI model evaluation work and built an authorization audit that tests whether AI agents act beyond what they're allowed to.

What the judges said

“The best on separating measurement from inference. The 80-case evaluation design, plus a metric that cannot be gamed by refusing, shows genuine epistemic discipline.”
Tom Leung · Judge
“Clearly offering human thought and judgement at a level well above AI. Solid pushback on AI recommendations.”
Forrest Corbett · Judge
View KyJahn’s Provn profile
Headshot of Prateek Pagare

3rd Place$100

Prateek Pagare

Data scientist who built an evaluation pipeline that tested 8+ AI models on quality, speed and cost at RentPrompts. Three went to production.

What the judges said

“The sharpest single piece of causal analysis in the group. Finding the selection bias, building a control group, and catching a hallucinated revenue figure in the board summary is excellent analytical work.”
Tom Leung · Judge
“It's far too common for people to key on outliers. He caught that a better baseline was needed, and available.”
Forrest Corbett · Judge
View Prateek’s Provn profile

Forward Deployed Engineer

The challenge: Step into a live AI freight-quoting pilot. Assess what the numbers really say, rank the failure modes by financial exposure, and give the customer’s leadership a recommendation they can repeat to their board.

Headshot of Vidula Sawant

1st Place$1,000

Vidula Sawant

Software engineer who builds and fixes C#/.NET and SQL systems that run warehouse and ERP operations. Earned the best average rank in the ladder.

What the judges said

“On video you reasoned beyond your documents. You left out a total dollar figure because ‘any total I wrote down would get forwarded and treated as a fact.’”
Rich Evans · Judge
“Excellent note to Dana. The TL;DR was direct and immediately answers what Dana needs. The use of AI was a clear example of an effective collaboration model.”
Keshav Puttaswamy · Judge
View Vidula’s Provn profile
Headshot of Nileet Savale

2nd Place$250

Nileet Savale

CTO and co-founder of Stamp’d, a travel compliance platform, where Nileet owns everything from architecture to deployment.

What the judges said

“The best articulation that spoke to an exec, non-technical audience while backing it with strong data analysis.”
Keshav Puttaswamy · Judge
“You delivered bad news directly: ‘the assistant isn’t winning extra loads yet.’ You found the variable that decides A versus B.”
Rich Evans · Judge
View Nileet’s Provn profile
Headshot of Liyan Cen

3rd Place$100

Liyan Cen

Applied AI engineer and former Amazon SDE who builds agentic systems on one rule: the agent proposes, independent evidence decides.

What the judges said

“When the AI moved your threshold from 70 to 80 with no new data, you asked ‘why 80?’ and kept 70 as provisional. You ranked failures by financial exposure, not frequency.”
Rich Evans · Judge
“The briefs were crisp and clear with strong reasoning. The video walked me through the recommendation with just the right amount of detail and data.”
Keshav Puttaswamy · Judge
View Liyan’s Provn profile

AI / Agentic Software Engineer

The challenge: Build a gate that turns customer emails into freight loads without costly errors. Debug an agent that got slower and dumber. Decide whether an agent that can spend real money is ready to ship.

Headshot of Ariel Agor

1st Place$1,000

Ariel Agor

Agent builder who has built a swarm of agents to execute tasks.

What the judges said

“Really nice job on systematically being a skeptic at all stages of the problem. Very grounded in adversarial use of competing models to check your own work.”
Jeff Kunins · Judge
“Removing any of 17 guards makes a test fail, which is real mutation testing. You caught the confused deputy and the falsy-zero ceiling bug.”
Ranga Komanduri · Judge
View Ariel’s Provn profile
Headshot of Jim Galvan

2nd Place$250

Jim Galvan

Backend engineer with 6+ years building .NET services, who builds open-source browser agents on the side.

What the judges said

“The best tool audit in the group: 13 defects grouped by failure class. The regression trick of replaying the same transcript and diffing token counts is deterministic. Clever.”
Ranga Komanduri · Judge
“I loved how you used different ways to explain all the challenges: code walkthroughs, animated presentations, and documents with your plan.”
Bhavesh Mehta · Judge
View Jim’s Provn profile
Headshot of Siddhanth Ramani

3rd Place$100

Siddhanth Ramani

Led Document AI for engineering drawings at Autonomous Machining, shipping the first working product in about 23 days.

What the judges said

“Proof-carrying extraction. The model must quote its source and code checks the quote appears verbatim. A cheap, deterministic confidence signal.”
Ranga Komanduri · Judge
“High conviction and an explicit methodology: deterministic approaches wherever possible for accuracy, speed and cost, and LLMs only where their advantages serve as advantages.”
Jeff Kunins · Judge
View Siddhanth’s Provn profile

Judged on the work. Nothing else.

Every challenge was built to fill a real skill gap in a role AI has changed. Judges scored what each candidate built and how they explained it.

The feedback above is quoted directly from the judges’ scorecards.

Meet the judges
  1. Build something real

    Candidates completed role-specific challenges based on real business problems.

  2. Show the thinking

    A video walkthrough of each decision, including where AI helped and where they pushed back on it.

  3. Ranked by AI leaders

    Judges ranked every submission in their ladder and wrote specific feedback on strengths and ways to grow.

Next steps

Hiring for AI-native roles?

These winners proved they can do the work before anyone asked for a resume. Meet them, or let us run a challenge for your open role and get a ranked shortlist of interview-ready candidates.

Hire with Provn

Want to be next?

Pick a challenge, build something real, and show employers how you think. Your work becomes your resume.

Take a challenge
Competition Rules · closed Nov 1, 2026

National AI Talent Draft and National AI Competition refer to the same competition.

Who can enter
U.S. residents, 18+, with one Provn candidate account. Full eligibility requirements apply.
How you're scored
Up to 2 attempts per Wave challenge; your best counts. Waves 1 to 3 add up.
Who advances
Top 10 per ladder (ties at 10th all advance) go to the judging panel.
Cost
Free. No purchase necessary.
Full official rules

The National AI Talent Draft is a skill-based competition run on Provn's Proving Ground platform across 8 role ladders, each a Ladder. Each Ladder contains 3 challenges released in waves. The 8 Ladders are:

  • AI Product Manager
  • AI / Agentic Software Engineer
  • Forward-Deployed Engineer
  • Data Analyst / Data Scientist
  • Product Designer (AI-native)
  • Growth / AI-Enabled Marketer
  • Account Executive / Solutions Engineer
  • Program Manager / TPM / AI Ops

This Competition is sponsored by Provn, Inc., located at 1100 Bellevue Way, Suite 8a, #823, Bellevue, WA 98004.

Eligibility

To be eligible to win a prize in the Competition, a builder (Entrant) must, at time of entry and at time of any prize award:

  • Be a legal resident of the fifty (50) United States or the District of Columbia, and be physically located within the United States at the time of submission.
  • Be at least 18 years of age or the age of majority in their state of residence, whichever is greater, as of the start of the Competition Period.
  • Have a valid Provn candidate account in good standing.
  • Not be an employee, contractor, officer, or director of Provn, Inc., or an immediate family member or member of the same household of such a person.
  • Complete at least one (1) of the 3 Challenges within a given Ladder to be eligible for that Ladder's prize.

Void where prohibited by law. Sponsor reserves the right to verify eligibility and to adjudicate any dispute at its sole discretion.

How to Enter

No purchase, payment, or fee of any kind is required to enter or win. Entrants compete by creating a Provn candidate account, if they do not already have one, and completing one or more of the 3 Challenges within a Ladder through the Proving Ground platform during the Competition Period.

Limit: one (1) Provn candidate account per person. Sponsor reserves the right to disqualify duplicate, fraudulent, or automated accounts and any Entrant who circumvents these Official Rules through multiple identities, bots, scripts, or other automated means.

Completing a Challenge uses the same Provn candidate account and profile system as the rest of the Provn platform. Profile data and visibility continue to work as described in the Data Use & Candidate Profile section below.

Ranking & Judging Methodology

Scoring

Each Challenge is scored against a rubric specific to that Challenge and Ladder. Initial scoring is performed by Sponsor's AI-assisted evaluation process and is final and binding for Combined Ranking and Top 10 Advancement, subject to Sponsor's discretion to review and correct manifest scoring errors. Selection of the Ladder Winner from among the Top 10 Finalists is made by the Judging Panel using the same rubric criteria by human judgment.

Combined Ranking

An Entrant's ranking within a Ladder is the straight sum of their best kept score on each of the 3 Challenges. A Challenge the Entrant does not attempt contributes zero (0). There is no averaging, weighting, or normalization across Challenges.

Advancement to Judging

For each of the 8 Ladders, the ten (10) eligible Entrants with the highest combined scores at the close of the Competition Period advance to judging. Ties at the 10th position all advance, so a Ladder's Top 10 may include more than 10 Finalists. No tiebreak narrows that group before judging. If fewer than 10 eligible Entrants complete at least one Challenge, the Ladder's prize will not be awarded.

Judging Panel & Winner Selection

The Finalists in each Ladder are reviewed by a Sponsor judging panel. The panel selects one (1) Ladder Winner by re-evaluating Finalists' submissions across all 3 Challenges against the disclosed weighted rubric criteria, applying human judgment and the same straight-sum method. A tie is resolved by the Judging Panel's discretion and discussion; its determination is final.

An Entrant may win no more than one (1) prize across all 8 Ladders. If an Entrant would otherwise win more than one Ladder, they receive the prize for the Ladder in which they placed highest, and the panel selects the next-preferred Finalist for each additional Ladder. Sponsor expects to announce Ladder Winners by November 9, 2026, or earlier.

Resubmission Policy

Each Entrant may submit up to two (2) attempts per Challenge.

Each attempt is scored independently. For each Challenge, only the higher of the Entrant's two scores is kept and carried into the combined sum. Attempts on different Challenges are independent.

If Sponsor verifies that an attempt was affected by a technical failure of the Provn platform, such as an outage or failed file upload due to a system error, that attempt will not count against the two-attempt cap and the Entrant may make a replacement attempt.

Attempts after the Competition Period, or after an Entrant exhausts their two attempts absent a Sponsor-acknowledged technical failure, will not be scored or considered.

Prizes

One (1) prize of $1,000 (USD) will be awarded to the Ladder Winner of each of the 8 Ladders, for a total of $8,000 (USD). An additional $2,000 is reserved for runner-up prizes, awarded by score and judging consistent with the rest of this Competition.

Prizes are non-transferable and no substitution is permitted except at Sponsor's sole discretion. Sponsor may substitute a prize of equal or greater value if the announced prize becomes unavailable.

Prizes constitute taxable income. Winners with total prize value from Sponsor of $600 or more in the applicable calendar year must complete IRS Form W-9 before disbursement, and Sponsor will issue Form 1099-MISC as required by law.

Disqualification, Fraud & Integrity

Sponsor may disqualify an Entrant, void a score or submission, or withhold a prize if it determines that an Entrant:

  • Submitted content generated through prompt injection, adversarial manipulation of Sponsor's evaluation systems, or an attempt to manipulate or reverse-engineer the scoring rubric or AI evaluator rather than genuinely complete the Challenge.
  • Created or used more than one Provn candidate account, or otherwise entered or attempted to enter using multiple identities, bots, scripts, or automated tools.
  • Submitted plagiarized, fabricated, or materially misrepresented work product.
  • Provided false or materially misleading eligibility, identity, or location information.
  • Otherwise violated these Official Rules or engaged in conduct Sponsor deems to be fraud upon or an unfair manipulation of the Competition.

Sponsor may disqualify anyone tampering with entry or Competition operation, violating these Rules, or acting unsportsmanlike or disruptively. Deliberately damaging the Proving Ground platform or undermining legitimate Competition operation may violate criminal and civil law, and Sponsor may seek remedies to the fullest extent permitted.

General Conditions

No Purchase or Payment Necessary

No purchase, payment, or fee is required to enter or win; participation is limited only by the eligibility and completion requirements stated above.

No Guarantee of Employment

Participation, submission, advancement, selection by the Judging Panel, or receipt of a prize does not constitute an offer of employment, a guarantee of employment, or an assurance of consideration for any position with Sponsor or a third-party employer. This Competition is a skills demonstration only.

Publicity

Except where prohibited, acceptance of a prize constitutes each winner's consent to Sponsor's use of their name, likeness, city or state of residence, and submitted work product for Competition promotional purposes without additional compensation, unless otherwise required by law.

Data Use & Candidate Profile

Completing a Challenge requires a Provn candidate account. That account includes a candidate profile, including name and contact information and, if provided, resume, work history, and education. Challenge submissions and scores are associated with that profile and are not anonymous.

This Competition does not change profile visibility. Employer visibility and interest in job services remain governed by the same account settings that apply to every Provn candidate profile. New accounts are private and not marked as interested in job services by default; an Entrant becomes employer-visible only by separately changing that setting. Personal information is used in accordance with Provn's Privacy Policy.

Limitation of Liability

By entering, Entrants agree to release and hold harmless Sponsor, its parent, subsidiaries, affiliates, and their officers, directors, employees, and agents from claims, injuries, losses, or damages arising from participation or acceptance, use, or misuse of a prize.

Right to Cancel, Modify, or Suspend

Sponsor may cancel, suspend, or modify the Competition or these Official Rules if fraud, technical failure, or another factor impairs Competition integrity, subject to applicable law.

Governing Law & Disputes

These Official Rules are governed by the laws of the State of Washington, without regard to conflict-of-laws principles.

Sponsor

Questions regarding these Official Rules may be directed to info@provn.co.