Vansh
"Turning ideas into working products." Responsible for model architecture files, training script configurations, pipeline construction, and backend environments.
TechnoTroob / AI + ML
Two people.
Two people.
Real builds.
One direction.
TechnoTroob is a two-person student-led AI/ML team building ideas into projects and taking them into hackathons.
IDEA
→
BUILD
→
PRESENT
→
COMPETE
→
LEARN
02 / The Team
The Core Duo
Two complementary builder roles. We construct local ML pipeline logic and build responsive interface interactions in parallel.
Divyansh Attrey
"Shaping ideas visually and communicating them effectively." Responsible for styling systems, UI workflows, presentation guidelines, mockups, and client stories.
03 / Methodology
How We Sprint
A fast-paced methodology crafted for high-stakes competition and experimental builds.
01 / Idea
Target Problems
We identify practical agricultural or automation issues that require localized AI/ML processing to solve.
02 / Build
Write Code
We assemble data structures, train local convolutional weights, and compile lightweight Python classification scripts.
03 / Present
Interface Mockups
We wrap model pipelines in clean, responsive web layouts so users can verify predictions in real-time.
04 / Compete
Hackathon Sprints
We enter hackathons. Strict 48-hour sprints force us to keep model setups lean and optimize execution speeds.
05 / Learn
Code Analysis
We analyze jury critiques and user interaction logs to upgrade our pipeline libraries for the next build.
06 / Iterate
Code Iteration
We reuse insights from prior sprints to optimize data pipelines, packaging improved code frameworks for upcoming events.
04 / What We Built
First Hackathon, First Win
CropShield AI
We built CropShield AI during our debut sprint at the AAROH 2026 hackathon. The system operates as a lightweight computer vision script designed to diagnose agricultural anomalies locally on edge setups, achieving 3rd position in AI/ML.
Rather than relying on high-bandwidth remote APIs, CropShield runs inference locally. This makes diagnostic reports instant and accessible even on farm sites without cellular networks.
Win Venue
AAROH 2026
Build Type
Computer Vision Model
05 / Core Toolchain
The Stack
The languages, libraries, and design systems we package to compile functional model setups.
Python
PyTorch
TensorFlow
OpenCV
NumPy
Next.js
JavaScript
HTML5 / CSS3
Tailwind CSS
React Native
Flutter
Git & GitHub
06 / Current Chapter
The story doesn't end
The story doesn't end
with the win.
Winning AAROH 2026 proved our builder approach. We are currently in a study and preparation cycle, optimizing data preprocessing pipelines for our next competition sprint.