Our Journey Through the ANNAM.AI Hackathon
FASALGEO: Field Analysis using Satellite Assessment of Land through Geospatial Environment Observation
Team Members:
Vaibhav Sharma (Leader), Shreya Khantal, Prasanna Saxena
Madhav Institute of Technology and Science, Gwalior
Mentors:
Dr. Soma Dhavala, Dr. Rajsekar Manokaran
About Us
We are three undergraduate students from Madhav Institute of Technology and Science, Gwalior. Vaibhav Sharma and Shreya Khantal are pursuing their B.Tech in Artificial Intelligence and Data Science, while also simultaneously earning a BS in Data Science and Applications from IIT Madras. Prasanna Saxena is completing his B.Tech in Artificial Intelligence and Machine Learning, complemented by a minor in Artificial Intelligence from IIT Ropar.
The Unexpected Beginning
It all began with a notification about the Annam AI Hackathon cum Internship Programme. As we were students studying Artificial Intelligence and Data Science, we were immediately drawn to this opportunity to work on agriculture-based solutions. However, reading through the eligibility criteria gave us pause – the program clearly prioritized “Final year UG or PG students (graduating in 2025 or 2026)” and “Students from agriculture-related institutions.”
Despite not fitting perfectly into these priority categories, we decided to take a chance. We carefully prepared our applications, highlighting our projects and knowledge in AI, ML and Data Science. We poured our hearts into showcasing what we could bring to the program.
When the email arrived confirming our selection, we were absolutely overwhelmed. That moment of being shortlisted despite the odds gave us not just validation but a precious opportunity to apply our technical skills to real-world agricultural challenges. We felt immensely grateful to the Annam team for looking beyond the formal criteria and recognizing our potential.
“Sometimes the greatest adventures begin with a simple leap of faith.”
The Qualification Challenge: Kaggle Competitions
When we received the selection email, our initial excitement was quickly tempered by the realization that our journey had just begun. To actually qualify for the internship programme, we learned we needed to excel in two Kaggle competitions against all other selected participants.
These competitions weren’t just formalities—they were intense challenges designed to test our technical capabilities and perseverance. The final deadline approached with alarming speed. As the clock ticked toward midnight on the submission day, we found ourselves still struggling to optimize our models.
The Final Moments: With just minutes remaining—literally at 11:55 PM, five minutes before the deadline—we finally discovered the parameter combination that yielded the best score. Those last moments were a blur of rapid implementation, verification, and submission. The pressure was immense, but so was our determination.
“True growth happens at the edge of your comfort zone—and sometimes, at the edge of a deadline.”
Our efforts paid off. We secured our place in the competition, officially qualifying us for the internship programme.
A New Perspective on Agriculture
The first week consisted of expert sessions designed to broaden our understanding of agriculture in India and the transformative potential of AI. These sessions weren’t just informative—they were eye-opening.
We discovered how cultural practices, regional weather patterns, and local economics all shape agricultural decisions. What might work as a solution in one region could fail entirely in another. These sessions transformed our thinking from “let’s build something cool with AI” to “let’s understand the real problems faced by farmers and create something meaningful.”
Meeting Our Mentor: A Turning Point
When we were assigned Dr. Rajsekar Manokaran as our mentor, we had no idea how valuable this relationship would become. Our first online meeting extended well beyond and we enthusiastically discussed potential problem statements.
What impressed us most was Dr. Manokaran’s approach. He didn’t just evaluate our ideas technically—he examined them through the lens of real-world implementation. For each problem statement we proposed, he asked tough questions and discussed them thoroughly. We discussed on various topics that included soil health, irrigation technologies, multi-spectral imaging, crop yield, monitoring, etc.
After much deliberation, we settled on addressing water optimization in agriculture. Water scarcity is a growing concern, and traditional irrigation methods often lead to wastage through over or under-watering. We proposed an AI-powered smart irrigation solution using multispectral imaging, in-soil sensors, and acoustic soil health scanners.
The Reality Check
Our initial euphoria quickly faced reality when we dove into research. We spent days searching for datasets related to multispectral imaging and soil sensors, reading countless research papers, and trying to understand the technical nuances of our proposed solution.
But we hit a roadblock. The datasets we needed weren’t readily available, and creating a rule-based system without proper data would lead to flawed results. We had envisioned something far too complex that wasn’t feasible with our available resources.
“In innovation, setbacks aren’t failures—they’re redirections toward better solutions.”
Pivoting with Purpose
This is where the Annam team truly shined. When we reached out about our challenges, they didn’t just offer suggestions—they assigned us an additional mentor, Dr. Soma Dhavala, who spent over an hour discussing our problem statement with us.
After this discussion, we made a critical pivot. Instead of water optimization, we shifted to “detecting crop types through remote sensing.” This wasn’t just a technical pivot—it was a pivot with purpose. We learned that this technology could help government agencies verify agricultural subsidies and disaster relief claims, reducing fraud and ensuring help reaches those who truly need it.
Real Social Impact: Dr. Dhavala explained how, during flood compensation programs, people sometimes falsely claim to have grown crops to receive government money. Our solution could provide objective verification through satellite imagery, ensuring fair distribution of resources.
What amazed us was the responsiveness of the Annam team and our mentors. On several occasions, Dr. Dhavala would respond at 2:30 AM with guidance and encouragement. This extraordinary level of commitment left us speechless and infinitely grateful.
The Research Marathon
The following days became a research marathon. We needed to understand:
- Remote sensing technologies and their limitations
- Available satellite data sources and their resolutions
- Machine learning approaches for crop classification
- How to handle the geospatial aspects of our solution
Both our mentors were incredibly supportive during this phase. These two brilliant mentors complemented each other perfectly – Dr. Manokaran providing the structured guidance and domain expertise we needed, while Dr. Dhavala brought technical insights and creative energy. Together, they created a support system that allowed us to achieve more than we could have imagined on our own.
“Behind every successful project is an invisible mountain of research papers, late-night discussions, and countless cups of coffee.”
Team Dynamics: Finding Our Rhythm
Our team quickly developed a working rhythm. We set daily milestones and held each other accountable. Every evening, we’d meet to share our progress, challenges, and insights.
What made this experience different from our university projects was the interdependence. In class projects, we could often work independently and combine our parts at the end. Here, our work was so interconnected that continuous communication became essential.
Late nights became the norm. There were moments of frustration when datasets weren’t cooperating or when complex geospatial operations wouldn’t render correctly. But there were also breakthrough moments—when a model finally started showing promising results or when we managed to visualize crop health on an interactive map—that made all the effort worthwhile.
Technical Hurdles and Supportive Solutions
As our project evolved, we encountered significant technical hurdles. Our crop classification models required substantial computational resources—far beyond what our CPU-based laptops could handle. Training complex models with satellite imagery and multispectral data demanded powerful GPUs and large storage capacities.
Game-Changing Support: This could have been a major roadblock, but the Annam team had anticipated this challenge and went above and beyond to support us. They provided us access to the TrueFoundry platform with high-performance computing resources including A100 and T4 GPUs—enterprise-grade hardware that would normally be far outside a student team’s reach.
The impact was transformative. Take our implementation of satellite image segmentation—what would take 30-40 minutes on our CPUs for even a small section completed in just minutes on TrueFoundry’s GPUs. This speed allowed us to test multiple variations, refine our approach, and drastically improve our segmentation results.
Learning Beyond the Curriculum
One of the most valuable aspects of this hackathon was how it pushed us beyond our academic comfort zones. As AI and DS students, we were comfortable with algorithms and model training. But this project demanded so much more:
- Understanding agricultural sciences and crop phenology
- Working with geospatial data and mapping technologies
- Learning about remote sensing and satellite imagery interpretation
- Grasping complex scientific formulas for vegetation indices and soil properties
- Designing intuitive interfaces for non-technical users
Every day brought new terminology and concepts we’d never encountered in our courses. What started as a hackathon became an immersive, interdisciplinary learning experience.
The FASALGEO Application Takes Shape
Gradually, our vision for FASALGEO—Field Analysis using Satellite Assessment of Land through Geospatial Environment Observation—took shape.
The application evolved beyond our initial crop detection focus to include:
- Comprehensive field analysis with crop health monitoring
- Soil composition and quality analysis
- Weather pattern assessment
- Anomaly detection with actionable recommendations
- Voice-based agricultural assistant for farmers who might struggle with text interfaces
Working with maps and geospatial data presented unique challenges. Unlike standard web applications where you control every pixel, maps have their own behaviors and constraints. Creating intuitive field selection tools, rendering crop health overlays, and ensuring the application remained responsive with large datasets required continuous problem-solving.
Looking Forward
The Annam AI Hackathon may have concluded, but our journey with FASALGEO feels like it’s just beginning. We see numerous possibilities for extending its capabilities and impact:
- Developing offline functionality for areas with limited connectivity
- Creating community features where farmers can share insights and best practices
- Expanding the crop disease detection capabilities to cover more regional varieties
Whatever the future holds for FASALGEO, we carry forward the most valuable lesson from this experience: technology is most powerful when it’s developed with empathy, understanding, and a genuine desire to improve people’s lives.
Reflections on Our Journey
As we finalized FASALGEO for submission, we reflected on how far we’d come from those initial days of uncertainty. The hackathon had transformed us in ways we hadn’t anticipated:
- We developed a deeper appreciation for the complexity of agricultural challenges and the potential of technology to address them
- We learned to approach problems from multiple perspectives—technical, economic, cultural, and practical
- We experienced firsthand how pivoting from an initial idea can lead to more impactful solutions
Most importantly, we realized that innovation isn’t just about creating new technology—it’s about creating meaningful solutions to real problems. And behind great innovation often stands great mentorship—something the Annam team provided in abundance.
“The true value of a hackathon isn’t measured by the projects it produces, but by the perspectives it transforms.”
The Annam AI Hackathon wasn’t just about building an application—it was about building ourselves as more thoughtful, collaborative, and purpose-driven innovators.