AI Conclave

The AI Hackathon is a multidisciplinary technical competition focused on applying Artificial Intelligence to solve real-world challenges across critical domains such as Environment, Healthcare, and Fintech. The event encourages participants to design innovative, ethical, and scalable AI-driven solutions that demonstrate both technical depth and societal impact.  With an emphasis on problem understanding, data-driven decision-making, explainability, and responsible AI practices, the hackathon provides a platform for students to translate theoretical knowledge into practical, real-world applications.  Introduction: The AI Hackathon aims to foster innovation, collaboration, and applied learning among aspiring technologists.  Participants will:  Identify and analyze real-world problem statements across predefined tracks.  Design and prototype AI-based solutions using appropriate datasets and tools.  Demonstrate feasibility through working prototypes or simulations.  Present solutions with a focus on impact, ethics, and scalability.  The event is structured to support participants of varying skill levels, from beginner-friendly approaches to advanced AI techniques.   Tracks & Problem Statements  Track 1: Environment  AI-Based Visibility Estimator in Fog, Dust, or Smog Problem Statement: Reduced visibility due to atmospheric conditions poses risks to transportation and public safety.  Objective: Develop an AI system that estimates visibility quality using limited inputs such as light source type, approximate particle density, and distance.  Expected Outcome: Classification of visibility conditions as Safe, Risky, or Unreliable.   Hyper-Local Pollution Forecaster Problem Statement: City-level AQI data lacks sufficient granularity for localized decision-making.  Objective: Build a predictive model to forecast pollution hotspots at a hyper-local level using historical climate data and satellite imagery.  Approach Levels:  Beginner: Time-series forecasting of AQI using open datasets.  Advanced: Correlating traffic density with pollution spikes and suggesting emission-reducing traffic strategies.   The “Zero-Waste” Vision (Computer Vision) Problem Statement: Inefficient waste segregation remains a major barrier to effective recycling.  Objective: Design an AI-powered computer vision system to classify waste items (E-waste, Organic, Plastic, Metal, Glass) and suggest correct disposal or reuse methods.  Approach Levels:  Beginner: Static image classification using pre-trained models.  Advanced: Real-time waste detection with carbon footprint estimation.   Track 2: Healthcare  CareQueue: AI-Assisted Hospital Queue Optimization Problem Statement: Inefficient scheduling and resource allocation increase patient waiting times.  Objective: Optimize outpatient scheduling and hospital resource usage to reduce average wait times.  Deliverables: Simulation or dashboard demonstrating scheduling improvements and predicted wait-time reduction.  MedAssist: Voice-Driven Medication Adherence Helper Problem Statement: Elderly patients frequently miss medication doses due to complex regimens.  Objective: Develop a privacy-first voice-based assistant for medication reminders, adherence logging, basic queries, and caregiver notifications.  Post-Operative Remote Monitoring Problem Statement: Post-surgical patients lack continuous monitoring after discharge.  Objective: Design an analytics system using wearable-device data to track recovery trends and flag potential health risks.  Approach Levels:  Beginner: Dashboard-based visualization and alerts.  Advanced: Predictive modeling to forecast deterioration trends.   Track 3: Fintech  MicroCreditScore: Inclusive Credit Scoring Problem Statement: Individuals without formal credit histories are excluded from financial systems.  Objective: Build an explainable credit scoring model using non-traditional, anonymized data sources while ensuring fairness and privacy.  InvoiceVerify: AI-Based Invoice Reconciliation Problem Statement: Small businesses struggle to reconcile physical invoices with digital records.  Objective: Develop a computer-vision-based system to extract invoice data, reconcile records, and flag discrepancies.  The “Invisible” Fraud Detector Problem Statement: Traditional fraud detection systems generate high false-positive rates.  Objective: Create a real-time fraud detection engine using behavioral and contextual data.  Approach Levels:  Beginner: Transaction classification using traditional ML models.  Advanced: Graph-based fraud detection with explainable outputs.    EVENT STRUCTURE  Round 1: Ideation and Prototype (Preliminary Round)  Submission Requirements:  Presentation (maximum 10 slides) covering problem definition, target users, AI approach, feasibility, expected impact, data sources, privacy considerations, and demo link.  A working prototype or demo (video, live demo, or mocked implementation).  Evaluation Criteria: Problem relevance, feasibility, innovation, and prototype clarity.  Timeline:  Submission window: 1 week  Evaluation and shortlisting: 3–5 days    Final Round: Complete Solution  Deliverables:  Deployed or runnable project with source code repository.  Architecture diagram and detailed system explanation.  Ethics and privacy section with bias mitigation strategies.  Technical report (2–3 pages).  Live demonstration and Q&A session.  Presentation Format:  8–10 minutes presentation  5–8 minutes Q&A    Eligibility: Open to students from any recognized engineering or technical institute.  Inter-college and inter-department teams are allowed.  Participants must carry a valid college ID.  No restriction on the number of teams per institute.  Team Rules  Team size: 1–4 members.  Each team must register with a unique team name.  A Team Leader must be designated during registration.  All official communication will be conducted through the Team Leader.   General Rules: Participants must adhere to ethical AI practices and data privacy norms.  Public or synthetic datasets are encouraged; sensitive data must be anonymized.  Plagiarism or reuse without attribution will result in disqualification.  Organizers reserve the right to modify rules or timelines.  Judges’ decisions will be final and binding.   Judging Criteria: Innovation and originality  Depth and correctness of AI implementation  Practical applicability and scalability  Explainability and ethical considerations  Robustness of the demo and presentation