🌾 Research Focus
India's crop breeding programs take 8–12 years to release a single improved variety, largely because genomic datasets across ICAR centres, state agricultural universities, and private seed companies remain siloed and underutilised. This research develops Federated Prior-Fitted Networks (Fed-PFN) — a privacy-preserving AI framework where institutions collaboratively train a shared national genomic prediction model without ever sharing raw data. By extending transformer-based Prior-Fitted Networks to multi-trait, multi-environment settings under federated learning, the system can predict crop yield, disease resistance, and climate adaptability directly from DNA markers before a single seed is planted.
Research Details
Federated Prior Fitted Networks for AI-Driven Multi-Trait Genomic Prediction in Indian Precision Agriculture
Federated Privacy-Preserving AI
Institutions train local models on private genomic data; only model weights are shared — never raw data.
Multi-Trait Genomic Prediction
Extending GPFNs to predict multiple crop traits simultaneously: yield, disease resistance, and climate resilience.
Multi-Environment Learning
Models generalise across diverse Indian agro-climatic zones — unlike current GBLUP methods.
Pre-planting Disease Prediction
Predict disease susceptibility from DNA markers before costly field trials, eliminating disease-prone lines early.
Key Research Keywords
Educational Background
Technical Skills
Core Skills
Programming & Tools
Data & ML Workflows
Selected Projects
National Child Rights Portal — NCPCR
Smart India Hackathon 2017 (Winning Project). Led development of the national-level digital portal for the National Commission for Protection of Child Rights.
Migrant Workers State Portal — Kerala
Smart India Hackathon 2025 (Winning Project). Architected and developed a state-level portal for migrant worker registration, tracking, and welfare scheme integration.
AI & ML Projects
Implemented supervised and unsupervised algorithms (linear regression, SVM, decision trees, random forest, KNN, K-means) and built end-to-end ML pipelines including preprocessing, feature engineering, training, evaluation, and hyperparameter tuning.
Achievements & Recognition
Teaching Assistantship
Contact Information
"Federated Learning is not just a technique — it's a philosophy of collaboration without compromise."