TwoStack - HackFlow(Build Your Own Dataset Challenge)

Inspiration

During TAMU Datathon 2026, we noticed an interesting problem. Around 600 participants were working inside the same hall, but pedestrian traffic was not evenly distributed. Certain walkways became noticeably more congested during lunch, dinner, and scheduled activities.

We wondered: What if we could use data collected from our own event environment to help participants navigate the hall more efficiently?

That question inspired HackFlow, an indoor navigation application that considers event schedules and congestion patterns when recommending walking routes.

What It Does

HackFlow helps participants navigate a large event hall by comparing the shortest walking route with a potentially less-congested alternative.

Key features include:

Interactive Hall Map: A navigable representation of the actual TAMU Datathon venue, divided into 15 walking sections.
Event-Aware Congestion Prediction: Estimates congestion based on the current event period and our location-specific dataset.
Automatic Time Detection: Uses the current time to identify the relevant event scenario.
Smart Route Comparison: Compares the shortest route with an alternative that accounts for estimated congestion.
Scenario Exploration: Allows users to explore how route recommendations change during different event activities.

How We Built It

We started by studying the physical hall layout and dividing the walking areas into 15 sections.

Our team, TwoStack, consists of three members: Rohit, Durgesh, and Aditya. Each member was assigned five sections of the hall.

We observed pedestrian activity and identified congestion patterns during six event scenarios, including regular hacking hours, lunch, dinner, and overnight activities.

We organized this information into a dataset containing 90 section-event records across 15 sections and six event periods. The numerical pedestrian counts are estimates based on our observed congestion patterns rather than individually measured traffic totals.

Technologies and methods used:

HTML, CSS, and JavaScript: Built the interactive website and navigation interface.
Excel and CSV: Organized the location-specific congestion dataset.
Graph-Based Routing: Represented walkable connections between sections and calculated alternative routes.
Event-Based Congestion Weighting: Adjusted route recommendations using estimated pedestrian traffic during different event periods.
GitHub Pages: Deployed the application as a publicly accessible website.

The dataset is central to HackFlow. Its location-specific congestion estimates influence the recommended routes, allowing the application to adapt to different event scenarios.

Challenges We Ran Into

Our biggest challenge was translating the physical hall into a connected digital navigation map.

Unlike conventional road navigation, our hall contains narrow walkways, multiple intersections, tables, and different entry points. We needed to represent these connections accurately so the application would recommend walkable routes.

Another challenge was accounting for congestion that changes throughout the event. Instead of treating every walkway as equally crowded, we designed the routing system to use scenario-specific congestion estimates.

We also focused on creating a straightforward, responsive interface that participants could use without additional instructions.

Accomplishments That We're Proud Of

Created an interactive navigation map based on the actual TAMU Datathon venue.
Organized location-specific congestion information across 15 sections and six event scenarios.
Built a working routing algorithm that compares walking distance and estimated congestion.
Integrated automatic time-based event selection.
Developed an interface that allows users to compare routes under different congestion scenarios.
Successfully deployed a functional, mobile-friendly website during the hackathon.

What We Learned

This project showed us that useful applications do not always require large external datasets or complex machine-learning models.

We learned how to turn observations from a physical environment into structured data, represent indoor walkways as a connected graph, and use event-specific information to support route recommendations.

We also learned the importance of distinguishing estimated congestion from directly measured traffic and accounting for uncertainty in predictions.

Most importantly, we learned how to identify a real-world problem, design a data-driven solution, and turn it into a working application within a limited timeframe.

What's Next for HackFlow

We would like to improve HackFlow by:

Incorporating live pedestrian counting to provide real-time congestion updates.
Collecting repeated measurements across more event periods to improve prediction accuracy.
Validating estimated travel times against actual walking times.
Expanding the navigation system to support larger and more complex indoor venues.

Eventually, HackFlow could be adapted for conferences, university events, exhibitions, and other crowded indoor spaces.

Our goal is simple: help people spend less time navigating crowded spaces and more time enjoying the event.