The Problem
We live in the age of deepfakes and AI-generated imagery. The ability to trust a photograph is becoming one of the most critical infrastructure problems of our time. Most detection tools analyze after the fact — too late.
The Approach
Hardware-first proof of authenticity — trust is embedded at the moment of capture, not verified later. A physical device combined with Web3's immutable ledger creates a chain of custody for visual truth that cannot be faked.
Why it matters
Journalism, law, security, and AI ethics all need this. It changes how courts handle evidence, how newsrooms verify images, and how humans navigate a world of synthetic media.
The Problem
Every AI assistant is reactive. It waits for a command and executes it. Siri, Alexa, Google — all the same pattern. Nobody has built AI that's genuinely proactive, emotionally aware, and contextually intelligent in daily life.
The Approach
Fusing Computer Vision (spatial awareness, child safety, object memory), NLP (natural conversation), Edge AI (on-device processing for privacy), and Behavioral Modeling (learning patterns over time) into one coherent, living system.
Why it matters
Most AI is built for utility. Sukku is built for connection. It remembers where you left your keys. It senses your mood. It learns your rhythms. It's what AI companionship should actually feel like.
The Problem
Every beginner ML project classifies 6 basic emotions. Real human emotional reality has 26+ complex states — contempt, awe, confusion, embarrassment — the ones that are hard even for humans to name.
The Approach
Deep learning system classifying 26 complex human facial emotions from images, with active research into model depth optimization, better activation functions, and attention mechanisms.
Why it matters
The scientific foundation for Sukku's emotional awareness. Built from first principles up, not from an API down.
The Problem
Healthcare and AI rarely meet at the point of care, especially in countries where medical resources are unevenly distributed. Remote diagnosis is still mostly a promise, not a product.
The Approach
Python-based computer vision system for wound analysis using image processing and OCR integration. Analyzes wounds from images — deployable on low-cost devices for remote diagnosis.
Why it matters
Not a CRUD app. Not a clone. A system with actual diagnostic potential that could help someone who doesn't have a doctor nearby.
The Problem
Renewable energy capture needs smarter hardware that adapts to its environment. Static solar panels waste capacity. Ocean environments are underutilized.
The Approach
Arduino-based floating solar panel with sun-tracking using sensors and servos, designed for renewable energy capture in ocean environments. Full firmware in C++.
Why it matters
Proves full-system thinking — from hardware sensor to firmware to real-world deployment. Most CS students have never touched a circuit board.