Dr.

AMBEDKAR INSTITUTE OF TECHNOLOGY

(An Autonomous institute affiliated to Visvesvaraya Technology University, Belagavi, Accredited by NAAC,
UGC with ‘A’ Grade) Near Jnana Bharathi Campus, Bengaluru – 560056

School of Computer Science & Engineering

Presents

ASTRAVA – Hackathon
Advancing Technology for a Sustainable Earth
Website: [Link] Hackathon Date: 6TH &7TH March 2026

Domains and Example Problem Statements for Hackathon

1. AI for Healthcare
Domain Description:

AI for Healthcare focuses on building clinically reliable, privacy-preserving, and explainable intelligence systems
capable of operating across highly heterogeneous medical data sources. These systems integrate multi-modal
foundation models, Generative AI, retrieval-augmented generation (RAG), and agentic AI architectures to
support real-time clinical decision-making, predictive medicine, and precision healthcare.
The domain emphasizes trustworthy AI—including fairness, uncertainty estimation, regulatory compliance
(HIPAA/GDPR), and ethical deployment—while enabling scalability across hospitals, rural clinics, and edge
environments. Modern healthcare AI increasingly leverages LLM-powered clinical assistants built with
LangChain and LangGraph, autonomous decision agents, federated learning, and on-device inference to deliver
actionable insights without compromising patient privacy.

Sample Problem Statements (SPS):

1. Multi-modal Clinical Foundation Models

Develop large-scale multi-modal foundation models that fuse MRI, CT, genomics, lab reports, and
unstructured clinical notes to predict neurodegenerative disease progression (e.g., Alzheimer’s, Parkinson’s)
several years in advance, with uncertainty quantification.
2. ICU Digital Twins with Agentic Intelligence
Build real-time ICU digital twins using streaming vitals, laboratory results, and EHR data, powered by
agent-based simulation and reinforcement learning, to predict and prevent sepsis, cardiac arrest, or multi-
organ failure.
3. Causal AI for Clinical Decision Support
Design causal inference and counterfactual AI models that estimate treatment effectiveness from
observational hospital data, assisting clinicians in evidence-based decision-making beyond correlation.
4. Federated & Privacy-Preserving Learning at Scale
Create federated learning frameworks capable of handling non-IID medical data, label noise, and
population drift across hospitals, while ensuring compliance with data privacy regulations.
5. Early Mental Health Crisis Prediction with Explainable AI
Develop early-warning AI systems for mental health crises using longitudinal behavioral signals, speech
patterns, wearable data, and clinical text, ensuring explainability, transparency, and ethical safeguards.
 6. Few-Shot AI for Rare Disease Diagnosis
Build AI systems for rare and orphan disease detection using few-shot, zero-shot, and transfer learning
methods where labeled data is extremely limited.
7. AI-Driven Drug Repurposing Platforms
Design drug repurposing pipelines that combine molecular graph learning, protein–ligand interaction
modeling, generative chemistry, and clinical trial data to identify new therapeutic uses for existing drugs.
8. Trust-Aware & Uncertainty-Aware Medical AI
Develop risk-aware AI models that quantify confidence, uncertainty, and potential harm before surfacing
recommendations to clinicians, improving trust and adoption in clinical workflows.
9. Edge AI for Low-Resource Healthcare Settings
Implement edge-deployable medical AI systems capable of real-time diagnostics (e.g., radiology, vitals
analysis) in rural or offline environments with limited compute and connectivity.
10. Regulatory-Compliant AI Governance Systems
Create automated AI governance platforms aligned with HIPAA/GDPR that continuously monitor bias,
fairness, explainability, and model drift, and generate compliance-ready audit reports.

2. AI for Smart Agriculture

Domain Description:
AI for Smart Agriculture leverages predictive analytics, deep learning, generative AI, and edge intelligence to
transform food production systems. By combining satellite imagery, IoT sensors, drone data, and historical
climate records, AI enables precision agriculture, climate resilience, and sustainable resource management.
Advanced techniques such as agent-based decision systems, reinforcement learning, and RAG-powered
advisory assistants help farmers and policymakers make data-driven decisions under uncertainty.

Sample Problem Statements (SPS):

1. Build climate-resilient crop yield forecasting models using decades of satellite, weather, and soil data,
incorporating uncertainty estimation and scenario simulation.
2. Develop AI-driven farm digital twins that simulate irrigation, fertilization, and pest-control strategies using
reinforcement learning agents.
3. Create self-supervised and foundation models for crop disease detection with minimal labeled data and
cross-crop generalization.
4. Design edge AI systems for pest and weed detection operating under low power, low bandwidth, and harsh
field conditions.
5. Build AI models to predict soil microbiome evolution and long-term soil fertility using graph learning and
time-series modeling.
6. Develop multi-agent reinforcement learning frameworks to optimize crop rotation strategies across
multiple seasons.
7. Create early famine and yield-collapse prediction systems using regional geospatial data and generative
forecasting models.
8. Build autonomous drone intelligence systems with onboard AI for adaptive crop monitoring and targeted
spraying.
9. Design explainable AI and RAG-based advisory assistants that translate complex predictions into farmer-
friendly recommendations.
10. Develop AI-powered supply–demand forecasting platforms integrating generative AI and agentic
workflows to reduce post-harvest loss and price volatility.
3. AI for Cybersecurity & FinTech
Domain Description:
This theme focuses on designing highly secure, adaptive, and scalable AI systems to protect financial
infrastructure and digital ecosystems. Techniques include graph AI, adversarial learning, large language models
(LLMs), RAG-based threat intelligence, and autonomous cyber agents capable of real-time reasoning and
response.

Sample Problem Statements (SPS):

1. Develop zero-day attack detection systems using self-supervised representation learning over raw network
traffic.
2. Build AI-driven cyber threat hunting agents that autonomously reason, plan, and adapt to unseen attack
patterns.
3. Design graph neural network models for real-time fraud detection on billion-scale transaction graphs.
4. Create adversarially robust AI systems resilient to evasion and poisoning attacks.
5. Develop cross-platform identity fraud detection systems integrating biometrics, behavior analytics, and
transaction data.
6. Build AI-powered AML systems that balance detection accuracy with regulatory explainability
requirements.
7. Design secure federated learning frameworks for banks without exposing sensitive data.
8. Develop predictive AI models for market manipulation detection across equities, crypto, and derivatives
markets.
9. Build blockchain + AI platforms for automated smart contract auditing using LLM-based code reasoning.
10. Create AI-powered cyber ranges using generative and agentic AI to simulate real-world attack–defense
scenarios.

4. AI for Sustainability
Domain Description:
AI for Sustainable & Smart Cities focuses on applying advanced artificial intelligence, generative simulation,
and multi-agent systems to address complex urban challenges related to climate change, infrastructure
efficiency, resource optimization, and citizen well-being.

This domain leverages urban-scale heterogeneous data—including traffic flows, energy consumption,
environmental sensors, satellite imagery, and demographic data—combined with LLM-powered RAG
dashboards, digital city twins, and autonomous decision agents. The objective is to enable data-driven urban
planning, real-time operational intelligence, and long-term climate resilience, while ensuring privacy, ethical
compliance, and scalability.

Sample Problem Statements (SPS):

1. City-Scale Digital Twins for Integrated Urban Planning
Build comprehensive city-scale digital twins that integrate traffic patterns, energy usage, climate variables,
and population dynamics. These AI-driven simulations should use agent-based modeling and generative
forecasting to evaluate policy decisions, infrastructure changes, and emergency scenarios before real-world
implementation.
2. Street-Level Urban Climate Risk Prediction Systems
Develop AI models that predict hyper-local climate risks such as flooding, heatwaves, and air pollution at
street or neighborhood granularity. Systems should combine IoT sensors, satellite imagery, and historical
climate data to provide early warnings and actionable insights for city authorities.
 3. AI-Driven Energy Grid Optimization
Create intelligent energy management systems that optimize power generation, distribution, and
consumption in real time. Using renewable energy forecasting, reinforcement learning, and demand-
response strategies, the system should reduce energy loss, balance load, and lower carbon emissions.
4. Robotic AI Systems for Waste Sorting and Recycling
Design AI-powered robotic systems that automatically classify, sort, and recycle municipal and industrial
waste. These systems should use computer vision, embedded AI, and robotics to improve recycling
efficiency and support circular economy initiatives.
5. Real-Time Water Quality Monitoring & Prediction Platforms
Build AI platforms that monitor water quality in rivers, reservoirs, and distribution networks in real time.
The solution should integrate IoT-based water sensors, satellite data, and hydrological models to detect
contamination, predict risks, and trigger alerts.
6. Multi-Agent AI for City-Wide Carbon Reduction
Develop multi-agent reinforcement learning systems that coordinate energy, transport, industry, and waste
sectors to optimize city-wide carbon reduction strategies. The AI should simulate trade-offs and recommend
policies aligned with sustainability goals.
7. Privacy-Aware AI Surveillance & Emergency Response
Create AI systems for public safety and emergency management that utilize cameras, sensors, and incident
data while preserving citizen privacy. Solutions must include ethical AI design, anonymization techniques,
and explainability to ensure trust and regulatory compliance.
8. AI-Powered Urban Mobility Orchestration Platforms
Build intelligent mobility platforms that coordinate public transport, electric vehicles (EVs), ride-sharing,
and autonomous vehicles. AI should dynamically optimize routes, reduce congestion, and improve
commuting efficiency using real-time data and predictive models.
9. Predictive Maintenance for Urban Infrastructure
Develop AI-driven predictive maintenance systems for roads, bridges, pipelines, and utilities. By analyzing
sensor data, images, and historical maintenance records, the system should predict failures early and optimize
repair schedules.
10. City-Scale Anomaly Detection & Resilience Systems
Design AI systems capable of detecting abnormal patterns across city infrastructure, such as cascading
failures in power, water, or transport systems. The goal is to predict systemic risks early and improve overall
urban resilience.

5. AI for SpaceTech
Domain Description:
AI for SpaceTech focuses on applying advanced AI, data science, and autonomous systems to challenges in space
exploration, satellite operations, Earth observation, space weather, and mission safety. These problems reflect
real-world constraints faced by organizations like NASA and ISRO, including limited data, extreme environments,
mission-critical reliability, and real-time decision-making.

Sample Problem Statements (SPS):

1. AI-Based Space Weather & CME Prediction

Develop AI models to predict solar flares and Coronal Mass Ejections (CMEs) using particle, plasma, and
solar imaging data to protect satellites, power grids, and astronauts.
2. Satellite Health Monitoring & Fault Prediction
Build AI systems that analyze telemetry data to detect anomalies, predict failures, and recommend corrective
actions autonomously.
 3. Space Debris Detection & Collision Avoidance
Design AI models to track space debris using orbital data and predict collision risks for satellites in low Earth
orbit.
4. AI for Earth Observation & Disaster Monitoring
Develop AI pipelines that analyze satellite imagery to detect floods, wildfires, landslides, or cyclones in near
real time.
5. Autonomous Mission Planning using AI Agents
Create agent-based AI systems that autonomously plan satellite operations or rover tasks based on mission
constraints and priorities.
6. Planetary Surface Classification & Rover Navigation
Build computer vision models to classify planetary terrain and assist autonomous rover navigation on the
Moon or Mars.
7. AI-Based Payload Optimization
Develop AI systems that dynamically decide when and what satellite data to capture under power, storage,
and bandwidth constraints.
8. Multi-Mission Satellite Data Fusion
Design AI frameworks that combine data from multiple satellite missions to generate higher-quality insights
for climate and space science.
9. Radio Signal Intelligence & Interference Detection
Build AI models to detect anomalies, interference, or signal degradation in spacecraft communication
systems.
10. Astronaut Safety & Risk Prediction Systems
Develop AI-based early-warning systems to assess radiation exposure, equipment risk, or life-support
anomalies during missions.

6. Open Innovation
Domain Description:
Open Innovation encourages unrestricted, high-impact problem-solving using AI, hardware, and intelligent
systems. This theme is intentionally broad to enable cross-domain innovation, allowing teams to combine
Generative AI, Agentic AI, IoT, robotics, edge computing, and data science to address societal, industrial, or
humanitarian challenges.
Sample Problem Statements (SPS):
1. Smart Disaster Alert & Response System
IoT sensors monitor floods, fires, or seismic activity, while AI models analyze data to generate real-time
alerts and response suggestions.
2. Edge AI Rural Healthcare Monitoring
A portable IoT-based health monitoring system using on-device AI for basic diagnostics in remote, low-
connectivity regions.
3. AI-Driven Resource Rationing System
IoT-enabled tracking combined with AI logic to ensure fair, transparent, and efficient distribution of food,
water, or fuel during shortages.
4. Climate Risk Intelligence Platform
Environmental IoT sensors collect long-term data, and AI models predict climate risks such as heatwaves,
pollution, or droughts.
5. Smart Infrastructure Health Monitoring
IoT vibration and stress sensors integrated with AI models to predict structural degradation in bridges and
public buildings.
 6. Intelligent Water Management System
IoT flow and leakage sensors with AI-based optimization to reduce water loss and improve supply efficiency.
7. Ethical AI Smart Surveillance System
IoT cameras with explainable and bias-aware AI, ensuring privacy, transparency, and auditability in public
monitoring systems.
8. AI-Powered Smart Agriculture Platform
IoT soil and climate sensors paired with AI-driven irrigation and crop health recommendations.
9. Industrial Worker Safety & Hazard Detection
Wearable IoT devices with AI models to detect unsafe conditions and trigger alerts for preventive action.
10. Smart Energy Monitoring & Optimization
IoT energy meters integrated with AI forecasting to reduce power wastage and optimize energy consumption.

For Queries / Contact

For any questions related to the hackathon, domains, or problem statements, participants can reach out to:
• Hackathon President: Mourya S
• Email: mouryas@[Link]
• Phone/WhatsApp: +91 7795009361 / +91 9113556098

Important Instructions & Guidelines:

1. Teams must submit original ideas and solution concepts. Plagiarism, fabricated claims, or copying
from AI tools, online repositories, or existing projects may lead to immediate disqualification.
2. The listed domains and sample challenges are indicative only, meant to convey the theme, scope,
and difficulty level of the hackathon.
3. Based on idea submissions, 60 teams will be shortlisted and invited to the offline finale hackathon
round.
4. For all domains except Open Innovation, shortlisted teams will receive an on-the-spot problem
statement during the offline finale and must build their solution within the event duration.
5. Open Innovation Domain – Special Rules:
• Teams must continue with the same problem statement/idea submitted initially; no on-the-
spot problem statement will be provided.
• Hardware-based solutions are permitted only under the Open Innovation domain.
• Teams developing hardware solutions must bring all required hardware components
themselves; no hardware will be provided.
6. The use of AI tools is allowed for assistance and brainstorming; however, the core idea, logic, and
explanation must be the team’s own.
7. Incomplete, vague, or low-effort submissions may not be considered for evaluation.
8. Shortlisting and evaluation will be based on innovation, clarity, feasibility, and alignment with the
hackathon objectives.
9. The organizers reserve the right to modify guidelines, evaluation criteria, or timelines, with due
notification.
10. The decision of the screening and judging committee will be final and binding.