Q-Buster

No more waiting in line - AI manages the queue for you!

Comment

About the Project: Q-Buster Agent

Inspiration

In India, queues are everywhere - hospitals, RTOs, temples, ration shops, and even at passport offices.
On average, an Indian spends months of their life just waiting in line.

We felt this daily frustration ourselves and wanted to ask:
“Can AI save people from wasting hours in queues?”

That question inspired us to build Q-Buster Agent - an AI system that predicts queues, manages bookings, and guides people in local languages.

What We Learned

While building this project, we learned:

  • How queueing theory works:

    • ( \lambda ) = arrival rate
    • ( \mu ) = service rate
    • Waiting time depends on ( \frac{\lambda}{\mu} ).

  • How to clean and use real-world queue datasets like:

    • Queue Waiting Time Dataset (IEEE DataPort) - structured queue logs with arrival/service times.
    • Hospital Wait Time Dataset (GitHub) - real-world OPD waiting durations.
    • Eye Smart EMR Outpatient Visit Data (India) - 3M+ visit records from an Indian hospital chain, capturing temporal footfall trends.

  • How IBM’s Granite Models + Time Series Foundation Models (TSFM) can forecast demand surges.

  • How to build multi-agent workflows with IBM’s Agent Development Kit (ADK).

How We Built It

We combined IBM’s open-source tech assets with real-world queue datasets (Kaggle, IEEE DataPort, GitHub, and medical repositories) to simulate and forecast Indian queue behavior:

  1. Data Preparation

    • Data preprocessing & pipelines -> Used IBM’s toolkits to clean and transform queue data (arrival times, service start/end, waiting times, queue lengths).
    • Cleaned hospital logs (ER wait times, appointments, EyeSmart OPD visits).
    • Converted them into time-series for forecasting.
    • Tools: IBM preprocessing toolkits.

  2. Prediction Models

    • Trained Granite + TSFM models to predict crowd density, average waiting time, and no-shows.
    • Example: If OPD gets 200 patients/day, our model predicts peak hours and waiting time distribution.

  3. Agents (using IBM ADK)

    • CrowdSense Agent -> predicts crowd levels from historical + real data.
    • SmartSlot Agent -> dynamically allocates or reallocates bookings when last-minute cancellations or missed appointments occur.
    • VoiceBuddy Agent -> sends voice/SMS reminders in local languages.
    • LoadBalancer Agent -> gives administrators live dashboards & suggestions.

Challenges We Ran Into

  • Data diversity -> Queue behavior differs by place. We solved this by combining multiple datasets.
  • Local language support -> Many citizens are not comfortable with English apps, so we built voice reminders in Indian languages.
  • Scalability -> Training on large datasets (like EyeSmart EMR) was heavy. We optimized using IBM’s efficient preprocessing and fine-tuning tools.
  • Fairness -> Ensuring slots are allocated fairly without bias toward early-bookers or those with faster internet. We tested fairness benchmarks with IBM tools.

Impact

  • Citizens -> Save hours of waiting, reduced stress.
  • Hospitals & Gov. offices -> Better management, smoother operations.
  • Rural inclusion -> Voice-first reminders make it accessible to everyone.
  • Society -> Greater fairness, fewer bribes/agents needed.

In short:
Q-Buster Agent turns India’s biggest daily annoyance, queues, into an opportunity for AI-driven efficiency, fairness, and inclusion.

Accomplishments That We’re Proud Of

  • We turned a very common Indian frustration (long queues) into a working AI agent solution.
  • We managed to use real Indian datasets (Data.gov.in) instead of just synthetic data.
  • Built a multi-agent workflow (CrowdSense, SmartSlot, VoiceBuddy, LoadBalancer) using IBM ADK that actually simulates how queues can be predicted and reduced.

What We Learned

  • How to clean and process messy public datasets from India (hospital OPDs and more).
  • The basics of queueing theory, which helps explain why waiting times shoot up when demand is high.
  • How to use IBM’s Time Series Foundation Models (TSFM) and ADK to build an agent system.
  • The importance of inclusivity in tech -> reminders in multiple Indian languages, SMS/voice for rural users, not just fancy apps.

What’s Next for Q-Buster

  • Expand from hospitals to RTOs, temples, ration shops, mandis- anywhere queues waste people’s time.
  • Add AI-powered fairness checks -> ensuring equal access regardless of privilege or digital skills.
  • Open up an API for admins so hospitals, temples, or govt offices can plug Q-Buster into their existing systems.  

Built With

Share this project:

Updates