InfiniSynapse Vibe Coding Hackathon 2026
Data Science Hackathon: From Rows to a Decision
A data science hackathon entry needs inputs, a retraceable result, and a next action. A notebook with no application around it is not the rewarded form.
By InfiniSynapse · Last updated October 2, 2026
1st Global Edition
Vibe CodingHackathon
Online 2026
Oct 10–31, 2026 · Pacific Time
Free · Global
$900 / $600 / $300
01
Education
Schools and training providers who cannot see learner engagement, course results, or how seats and tutors are used.
Question Which courses lose learners before they finish?
02
Retail & Replenishment
Store operators deciding when to reorder and which items matter first.
Question What is about to stock out this week?
03
Hit-product Analysis
Merchandisers who need the reason a product is selling, not only a bestseller list.
Question Why did this product jump, and should we reorder it?
04
SMB Operations
Operators of a small business whose sales, costs, and orders live in separate places.
Question Which orders are eating margin?
05
Growth for Operators
A small team that cannot see which channel, conversion step, or retention pattern is worth their time.
Question Which channel should we spend the next week on?
06
Open Industry
Any operator you can name, in a setting the other tracks do not cover.
Question What decision does this operator repeat every week?
TL;DR
A data science hackathon entry for the InfiniSynapse Vibe Coding Hackathon 2026 is a working web, mobile, or desktop app. InfiniSynapse hosts the event. It is online and free. The build window is October 10, 2026, 12:00 AM through October 31, 2026, 11:59 PM Pacific Time. In that window Pacific Time is daylight time, UTC−7. The open is 07:00 UTC on October 10. The close is 06:59 UTC on November 1. Keep the clock in Pacific Time even if you build in another zone.
The entry needs inputs, a retraceable result, and a next action. A notebook with no application around it is not the rewarded form. Slide-only pitches, fake JSON standing in for a real path, and a generic chat box on a spreadsheet are outside the rewarded set too. Six tracks are starting points for one registration: Education, Retail & Replenishment, Hit-product Analysis, SMB Operations, Growth for Operators, and Open Industry. The prize pool is $3,000: one prize of $900, two of $600, and three of $300. Judging is real-world scenario value 40%, technical completeness 35%, and innovation 25%. GitHub stars are not in the score. William Zhu, Co-founder of InfiniSynapse, is a confirmed judge. The event page has not published age, region, or team-size rules. Payout timing, taxes, judging dates, a winner date, and capacity are not published. A submission deadline separate from the build window is not published. Register only on Luma.

What a data science hackathon entry includes
This page treats analysis as something a reviewer can run, not a stack of cells they have to trust on faith. IBM's topic page on data science describes the field as using scientific methods and algorithms to draw insight from data. The Wikipedia article on data science surveys the same public term, from collection through interpretation. Neither page is the rulebook. They only fix the vocabulary. The data science hackathon rule on this site is a working app around that chain.
Inputs a reviewer can see
Name the file, the fields, and the question before you name a model. A course question needs a course id and a completion or attendance field. A replenishment question needs on-hand units, recent sales, and lead time. A margin question needs revenue and cost on the same order line. If those columns are missing, say so on the screen. A data science hackathon demo that prints a confident sentence with no columns under it is not complete.
Label synthetic rows as synthetic. The event page has not published a shared dataset. Use rows you are allowed to use. A one-line note in the app can say whether the file is real data you may show or a set you built for the demo. That note is part of the path.
A result you can retrace
A reviewer should repeat the result: same inputs, same question, a next step they can read without your commentary. The Wikipedia article on statistical inference explains how a sample supports a claim about a larger set. You do not need a theorem in the demo. You need the comparison the screen actually used, written where the operator can see it. The data science hackathon result is that comparison plus one action, not a hidden score.
Fields before a sentence
Put the fields above the sentence they support. "Reorder this SKU first" should sit next to on-hand units, recent sales, and lead time. "Contact this cohort" should sit next to the completion gap. If a filter changes the rank, show the filter. Do not bury the inputs in a download the reviewer has to open off to the side.
The next action on screen
End on a verb the operator already uses. Reorder, follow up, feature, watch, or review the order. A rank with no verb is a chart, not a decision. The Vibe Coding Guide is the product guide for wiring that verb to a screen. Register only at the official Luma listing. This page does not publish a Discord invite.
Why a notebook alone is not the brief
A notebook can be how you explore. It is not the rewarded form when no application wraps it. The reviewer has to open a client, supply or select rows, and see the action. A slide that describes the notebook does not replace that client.
The rewarded form is an app
Ship one client. Web, mobile, or desktop all qualify. You do not need a native build for every platform. Pick the screen the operator already uses on a Monday morning. A data science hackathon entry that lives only in a slide file never reaches that Monday.
What stays outside the score
Slide-only pitches, notebooks without an app, fake JSON standing in for a path, and a generic chat box on a spreadsheet are outside the rewarded set. Fake JSON is a file that pretends a calculation ran. A chat box that answers any question about any sheet has no operator and no next step. Leave both out.
How a chart earns its place
The Wikipedia article on data visualization describes charts as a way to see structure in data. A chart belongs here when it changes the action. A bar of stockout risk next to an order priority earns its place. A decorative plot that does not change the click is optional. The data science hackathon chart is evidence for the verb, not a second product.
A practical check before you call the result done: pick the top row and recompute it by hand. If the screen says a course is first because completion is 42% against a 70% peer figure, those two figures should be on the page. If a SKU is first because on-hand units cover fewer days than lead time, show units, recent sales, and lead time as three separate numbers. A fused score with no parts is harder to retrace than the parts themselves.
Watch the grain. One row per course is a different question from one row per learner. One row per SKU is a different question from one row per store per day. Say which grain you used in the empty state and again above the table. Mixing learners and courses in the same list, or stores and SKUs in the same list, makes the verb ambiguous. State the grain in a short line the reviewer can read without opening your notes.
Partial files need an honest branch. If lead time is missing, do not silently invent a default of seven days. If you do use a default, print it and let the operator change it. If cost is missing, do not imply a margin. Show the gap and hold the action for that row, or ask for the column. That branch is part of technical completeness, in the same way the rank itself is.
Tracks as analysis scenes
The six names are scenes for one registration, not six contests. A model may summarize, rank, or draft the reason. The operator and the action stay specific. If the center of the demo is the model itself, read what an AI build still has to ship. For the app form this analysis has to wrap, read what a reviewer can open and try.
Six starts, one signup
Register once. Then pick Education, Retail & Replenishment, Hit-product Analysis, SMB Operations, Growth for Operators, or Open Industry. The event page for this October build lists the user, the data, the decision, and the demo for each name. A data science hackathon entry can cover one of those rows end to end.
| Track | Question the app can answer | Action on screen |
|---|---|---|
| Education | Which courses lose learners before they finish? | A follow-up for the worst drop-off |
| Retail & Replenishment | What is about to stock out this week? | An order priority |
| Hit-product Analysis | Why did this product jump, and should we reorder? | Restock, feature, or watch |
| SMB Operations | Which orders are eating margin? | The orders to review first |
| Growth for Operators | Which channel should we spend the next week on? | One focus, with the conversion gap |
| Open Industry | What decision does this operator repeat every week? | That decision, the reason, and the action |
An illustrative education direction
This is an illustrative direction, not a customer case study, and it does not promise a product feature beyond the public docs. A coordinator opens the app, selects a labeled roster, and asks which courses lose learners before they finish. The screen orders courses by drop-off and suggests a follow-up for the worst gap. Completion and attendance stay visible so the coordinator can disagree with the rank.
What the screen should show
Show the question, the roster fields, and the follow-up. If a term filter or a program filter changes the order, show that control. Do not hide the comparison behind a single chat bubble. Another illustrative direction, still not a case study: a buyer loads inventory and sees SKUs ordered by stockout risk plus an order priority. The pattern is the same in both scenes. Inputs stay on screen, the result can be retraced, and the verb is specific.
Score, cash, and the October clock
The score does not award the longest method name. It awards a situation a reviewer recognizes, a build they can run, and an approach that is more than a stock chart with no decision.
The three published weights
Real-world scenario value is 40%. Technical completeness is 35%. Innovation is 25%. Scenario value is the operator and the weekly decision. Completeness is the running client, the visible inputs, and the retraceable result. Innovation is a useful twist on that path, not a new chart type with no verb. Tie-break rules are not published. GitHub stars are not part of the score. The data science hackathon weights stay at those three numbers.
The cash table
The pool is $3,000. One first prize is $900. Two second prizes are $600. Three third prizes are $300.
| Place | Award | Seats |
|---|---|---|
| First | $900 | 1 |
| Second | $600 | 2 |
| Third | $300 | 3 |
Timing the page does not print
Payout timing and tax handling are not published. Do not budget a pay date from this guide. Capacity is not published either. The seat counts above are prize counts, not a cap on registration.
The build window
Build from October 10, 2026, at 12:00 AM Pacific Time through October 31, 2026, at 11:59 PM Pacific Time. A submission deadline separate from that window is not published. Judging dates and a winner announcement date are not published. Plan the demo inside the window you can already read.
Access and the blanks
The signup path is one listing and one product guide. Everything else that builders often assume is still blank until the event page prints it.
One listing for registration
After you RSVP on Luma, open the guide and pick one track. Write the question in a single sentence before you design screens. Then list the fields that sentence needs. If you cannot name those fields, the comparison has nothing honest to read. Build the empty state first, so a reviewer who arrives with no file still sees what to upload and how it should be labeled. Run the question on a small labeled set and check that the action still points at a row the reviewer can find. Save that check for the last evening of the window, in Pacific Time, so a timezone miss does not eat the demo.
Lines the event page leaves open
The event page has not published age, region, or team-size rules. Judging dates and a winner date are not published. William Zhu, Co-founder of InfiniSynapse, is a confirmed judge. A longer biography is not provided here. Until the event page fills a blank, that blank is not a rule you can invent. Use the published window, the three weights, and the six track rows. Those are the data science hackathon facts you can plan against today.
Use the three weeks as a sequence. In the first week, lock the operator, the question, and the columns. In the second week, make the client run on a small labeled file, including the empty state and the missing-column branch. In the last week, rehearse the top-row check in Pacific Time and cut any control that does not change the verb. A second chart that does not change the action can wait. A second platform can wait. The reviewer needs one client that opens, one file shape they understand, and one next step they can disagree with because the inputs are still on screen.
A data science hackathon question should fit in one sentence an operator would ask this week. A data science hackathon column list should be the fields that sentence needs, not every export on the laptop. A data science hackathon client should open before any model call is impressive. A data science hackathon empty state should tell the reviewer what file shape to bring. A data science hackathon missing field should stop the run with a plain message. A data science hackathon result should point at a row the reviewer can find. A data science hackathon chart that does not change the verb can wait. A data science hackathon synthetic file should say synthetic on the screen. A data science hackathon notebook with no client around it is not the rewarded form. A data science hackathon track is still one of the six starting points. A data science hackathon window is still October 10–31, 2026, Pacific Time. A data science hackathon score is still 40, 35, and 25. A data science hackathon star count is still not a score. A data science hackathon signup is still the Luma listing. A data science hackathon payout date is still unpublished. A data science hackathon age rule is still unpublished. A data science hackathon region rule is still unpublished. A data science hackathon capacity line is still unpublished. A data science hackathon winner date is still unpublished. A data science hackathon judging calendar is still unpublished. A data science hackathon separate submit deadline is still unpublished. A data science hackathon demo should still be one client a reviewer can open. A data science hackathon figure should still trace back to a column on the screen.

Questions about the analysis bar
Does a notebook qualify?
A notebook without an application around it is not the rewarded form. Exploration in a notebook can inform the app. The reviewer still needs a web, mobile, or desktop client they can open and try.
Do I need a shared dataset?
The event page has not published a shared dataset. Use rows you are allowed to use, and label synthetic rows as synthetic. Show the fields the result depends on.
Are stars part of the score?
No. GitHub stars are not part of the score. The weights stay at 40% for scenario value, 35% for technical completeness, and 25% for innovation.
Who judges?
William Zhu, Co-founder of InfiniSynapse, is a confirmed judge. Judging dates and a winner date are not published. The confirmation on this page stops at that name.
Are age and team size published?
The event page has not published age, region, or team-size rules. The steps you can act on today are the Luma RSVP, the product guide, and one track you can demo inside the October window.