IndiaAI Impact Gen-AI Hackathon Results

Celebrating Innovation in Generative AI for India

We are thrilled to announce the outstanding results from our 4-week-long hackathon dedicated to developing innovative generative AI solutions specifically designed for the Indian context. The level of enthusiasm, creativity, and technical excellence displayed by participants throughout this journey has been truly inspiring, culminating in exceptional ideas, project proposals, and finally solution submissions that address critical challenges across all three competition tracks.

Participation Highlights

Following our launch on August 14th, we were overwhelmed by the tremendous response from the student community. We received 3292 registrations from students representing 100 institutes across India, demonstrating the widespread interest in AI innovation among India's youngest minds.

Competition Journey

• Phase 1: 637 project proposals were received, showcasing diverse and creative approaches to real-world problems covering ideas around “Agriculture”, “Smart Governance”, “Health Care”, “Personal Assistant”, “Safety”, “Automation”, “Culture”, “Legal & Compliance”, “Education” etc. 303 proposals were accepted for Phase-2.


• Phase 2: 60 teams submitted their working solutions which were considered for final evaluations. These submissions represented the most promising and well-implemented, well-executed concepts.

Evaluation Process

Each submission underwent rigorous assessment by distinguished subject-matter experts from our IBM Research and IISc teams. Our comprehensive evaluation framework considered multiple critical dimensions:

• Technical Completeness - Thoroughness of implementation and functionality
• Practical Viability - Real-world applicability and usability
• Data Excellence - Quality and relevance of curated datasets
• Relevance - Alignment with current needs and challenges
• Potential Impact - Capacity to create meaningful change
• Documentation Quality - Clarity and comprehensiveness of technical documentation
• Scalability - Potential for growth and widespread adoption

Through this rigorous process, we are proud to announce our Top 3 winners who exemplified excellence across all evaluation criteria.

And the Winners Are...

After careful deliberation, the judging panel has selected the following teams for their exceptional work.

Track-1: Crop identification using satellite data

• First Place: Team “SAY” (Project ID T1:82)
  Team members:
  • Subhojit Mandal
  • Atin Chowdhury
  • Yashwardhan Singh Chouhan
  The submission performed fairly well by introducing various loss functions that work effectively on sparse datasets, and it clearly demonstrated how an ensemble of loss functions can be applied to fine-tune a geospatial foundation model for the crop identification task. This submission also curated additional feature data (Optical + SAR).
• Second Place: Team Maverick (Project ID: T1:95)
  Team members:
  • Soumabrata Bhowmik
  • Arnit Maiti
  • Ankush Chakraborty
  The contribution focused on how fine-tuning of geospatial data can be performed to handle sparse label data using a weighted combination of two well-known loss functions - Custom composite loss (focal + dice).
• Third Place: Team Crop_FusionFormer (Project ID: T1:40)
  Team members:
  • Kabilan Ganesan
  • Kalika Dakshini
  The submission curated SAR data and integrated it with the shared optical images to fine-tune the Prithvi model. Attempted Prithvi + SAR fusion approach.

Track-2: Short-term energy load forecasting

• First Place: Team “VISA” (Project ID: T2:7)
  Team members:
  • Sahana Ganeshvel
  • Vijaybala Mahalingam
  This solution combines the strengths of state-of-the-art forecasting models, such as the simple Hour-of-Day (HoD) baseline, TinyTimeMixer (TTM) with its strong zero-shot performance, and the LightGBM Regressor, enhanced through ensembling and stacking techniques to achieve a balance between performance and robustness against outliers.
• Second Place: Team “Spandan Chavan1” (Project ID: T2:77)
  Team members:
  • Spandan Chavan
  This solution adopts a foundation-model-as-feature approach: it combines Chronos’s zero-shot quantiles as rich meta-features with simple hand-crafted features, stacked through a lightweight Ridge regressor, to yield a robust and fast ensemble solution.
• Third Place: Team “Arisha” (Project ID: T2:27)
  Team members:
  • Arisha Rizwan
  This solution is based on a weighted ensemble of gradient boosting models, such as LightGBM and XGBoost, combined with features like holidays and various statistical attributes.

Track-3: Agentic AI applications for real-world issues

• First Place: Project "AyushmanAI-The Intelligent Medical Assistant" (Project ID: T3:365)
  Team members:
  • Aditya Puri
  This project submission show cases a working Health Care automation streamlining patient-doctor workflows for appointment scheduling, diagnosis documentation, and follow-up. It uses Langflow backend (integration with AstraDB). The solution has a potential use-case scope of deployment within Government Hospitals. The underlying solution can have a greater outreach to mass population with Indian Languages Support.
• Second Place: Project "StartupSarthi CA Agent" (Project ID: T3:98)
  Team members:
  • Vedika Sardeshmukh
  • Dnyaneshwari Pawar
  • Sakshi More
  This project showcases Virtual chartered accountant agent for Micro, Small, and Medium Enterprise for financial automation and compliance. The underlying solution showcased use of IBM Watsonx Orchestrate ADK to build CA Agent.
• Third Place: Project "WanderPal travel advisor" (Project ID: T3:29)
  Team members:
  • Aryan Talikoti
  • Sriram Kulkarni
  • Hemang Shrikar
  This solution presents an AI-powered travel companion featuring a complete full-stack architecture. The backend leverages the Langflow framework, with custom components seamlessly integrated to deliver a fully functional prototype implementation.

Acknowledgment

We extend our heartfelt gratitude to every participant who contributed their time, talent, and passion to this hackathon. Your dedication to innovation and commitment to developing solutions for India's unique challenges has made this event a remarkable success. The quality of submissions and the collaborative spirit demonstrated throughout the competition truly reflects the exceptional potential of India's emerging tech talent.

Vote of Thanks

Our heartfelt thanks go to the exceptional judging panel from IBM Research and IISc, who dedicated their valuable time and brought their deep expertise to the evaluation process. We are also immensely grateful to our webinar speakers, who not only delivered insightful sessions but remained available to support participants with guidance and answers throughout the entire hackathon period. Your combined efforts were the cornerstone of this event's outstanding success.

What's Next?

Congratulations to all our winners! We will be reaching out to the project leads of our top three teams in the coming days to discuss next steps, including prize distribution and exciting opportunities for continued mentorship. Additionally, all teams that submitted valid solutions will receive certificates of participation in recognition of their valuable contributions.