Our Journey to Smart India Hackathon 2025

Smart India Hackathon is one of India's largest innovation challenges — thousands of teams from colleges across the country competing to solve real-world problems posed by government ministries and organizations. In 2025, our 6-member team took on the challenge, and what followed was an intense, sleepless, and incredibly rewarding journey.

The Problem Statement

We picked a problem statement around loan utilization analysis — banks and financial institutions needed a way to verify whether loans disbursed for specific purposes (like agriculture equipment, housing, or education) were actually being used as intended. Manual verification was expensive, slow, and often unreliable.

We saw an opportunity to solve this with technology.

Building Nidhisetu

We named our solution Nidhisetu — a bridge (setu) for financial resources (nidhi).

The core idea was simple but powerful: build a React Native mobile application that field agents could use to photograph and document assets purchased with loan money, and then use AI to analyze and verify whether the assets matched the loan purpose.

The Tech Stack

• React Native with TypeScript for the cross-platform mobile app
• Ollama AI models for on-device asset recognition and analysis
• Node.js backend for data aggregation and reporting
• Real-time sync with offline-first architecture for rural areas with poor connectivity

Key Features We Built

1. Asset Photography & Geotagging — Field agents could capture photos with automatic GPS tagging, ensuring the asset was at the declared location.

2. AI-Powered Asset Analysis — We integrated Ollama models to analyze captured images. The AI could identify whether a photographed item (say, a tractor or construction materials) matched the loan category.

3. Dashboard & Analytics — A web dashboard for bank managers to track loan utilization across regions, flag suspicious cases, and generate compliance reports.

4. Offline-First Design — Since many loan recipients are in rural India, the app worked offline and synced data when connectivity was available.

The Hackathon Experience

Internal College Round

Before SIH, we had to clear our college's internal selection. We presented Nidhisetu at our institution — Government College of Engineering, Kalahandi — and were selected to represent our college at the national level. That validation gave us the confidence to push harder.

Preparing for the Finals

The weeks leading up to SIH were intense. We split responsibilities:

• Two of us focused on the mobile app UX — making sure the capture flow was smooth enough for non-technical field agents
• One teammate handled the AI pipeline — fine-tuning the Ollama model for asset categories relevant to Indian loan schemes
• Two worked on the backend and dashboard — building the reporting system that would actually be useful for bank officers
• I coordinated the overall architecture and worked on integrating all the pieces together

We iterated fast. Every evening we'd demo to each other, find gaps, and fix them overnight.

The Grand Finale

Walking into the SIH venue was surreal. Teams from all over India, each working on completely different problems — from healthcare to agriculture to defense. The energy was electric.

We had 36 hours of non-stop building, presenting intermediate progress to mentors and judges, refining our solution based on feedback, and preparing the final demo.

Some moments that stand out:

• The 3 AM breakthrough — Our AI model kept misclassifying certain agricultural equipment. At 3 AM, one of my teammates figured out the preprocessing was stripping color channels that were crucial for distinguishing between equipment types. A quick fix and suddenly our accuracy jumped.

• The mentor feedback — One of the industry mentors told us, "This is something banks would actually pay for." That moment made all the sleepless nights worth it.

• The final presentation — We had 8 minutes to convince the judges. We opened with the scale of the problem (₹15+ lakh crore in outstanding loans in India), showed a live demo of the app working end-to-end, and closed with our roadmap for scaling.

The Result — Finalist

We secured a Finalist position among 30,000+ participating teams nationwide. While we didn't take the top prize, reaching the finals of SIH was a massive achievement for our team.

More importantly, the process taught us things no classroom could:

• How to build under extreme pressure — 36 hours is not a lot of time. Every decision had to be fast and deliberate.
• How to work as a real team — Not a group project team, but a team where everyone owns a piece and trusts each other.
• How to present technical work to non-technical audiences — The judges included bureaucrats, bankers, and policymakers. We had to translate our tech into their language.
• How to handle setbacks — Things broke. APIs failed. The AI gave wrong results. But we kept going.

What's Next

Nidhisetu didn't end at SIH. We're continuing to develop it — improving the AI models, adding more loan categories, and exploring partnerships with regional banks for pilot testing.

The Smart India Hackathon was a turning point for me as a developer. It proved that building something meaningful isn't just about writing code — it's about understanding a problem deeply, working with people who complement your skills, and having the persistence to see it through.

If you're a student thinking about participating in SIH — do it. You won't regret it.

Feel free to check out Nidhisetu or connect with me on LinkedIn if you want to chat about hackathons, full stack development, or anything tech.