Flagship TIC Project
SARADHI AI: TARA — Applied Intelligence Product Line
TARA is SARADHI AI's comprehensive product suite for generative, agentic, edge, cognitive, and physical AI. It is designed to modernize teaching, assessment, healthcare diagnostics, aerospace modeling, and community service.
Funding Scheme:
Technology Innovation Centre (TIC)
Principal Investigator:
Dr. T. Venkatesh
Active Period:
December 2025 – June 2027
Delivery Modality:
Data-as-a-Service & Software-as-a-Service (DaaS / SaaS)
Lab Location:
AI & Robotics Lab, VV126, Vidyut Vihar, SASTRA
TARA ARCHITECTURE SUITE (19 WORKS)
Cloud Kubernetes Tier
SARADHI Web Platform · SARA Polls · AI Assessment Engine
Foundation & Reasoning Models
KALI / SARAS 1.5B · AgriSense · MediReasoning · ChessGambit
Edge Hardware Tier
SchoolBox (Jetson Orin NX) · CADAI Drone · Audio Quiz Spark
Ph.D. Research Fellowships
Bone Age Assessment from Pediatric X-Rays
Scholar: Sivaranjani R · Supervisor: Dr. B. Santhi
Non-invasive bone age assessment using deep learning on pediatric hand radiographs:
- ✓BoneSight: Hand localisation, MaxViT with residual CBAM attention, gender-aware fusion, SVR, and Eigen-CAM visual explanations.
- ✓BoneScale: Principal-axis bone geometry and carpal pixel areas fused with deep features, evaluated via XGBoost.
- ✓Both methods unified into the standalone CAMBAA clinical GUI.
Breast Cancer Detection from Thermal Images
Scholar: Abirami M · Supervisor: Dr. B. Santhi
Radiation-free screening aid classifying breast thermograms from the DMR-IR cohort with explainability:
- ✓U-Net breast ROI segmentation comparing performance before and after region extraction.
- ✓Radiomic intensity, texture, shape, GLCM and LBP features fused with EfficientNet features, reduced by PCA.
- ✓Classification via SVM, Random Forest, XGBoost and LIME region attributions.
| Research Track | Student Intern | Faculty Mentor | Validated Deliverable | Status |
|---|---|---|---|---|
| Breast Cancer Classification (Thermal Images) | Abinaya V (M.Sc. Data Science) | Dr. B. Santhi | 100+ evaluations across handcrafted, deep, fused & graph features; best fused F1 67.89% | Evaluated |
| Deep Learning for Osteosarcoma Detection | Kavya D (MCA) | Dr. R. Nithya | CNN–Swin Transformer app on Hugging Face; Graph Neural Network enhancement | In Progress |
| Explainable Hybrid AI for Brain Stroke CT | Devadharshini M (M.Sc. CS) | Dr. R. Rajakumar | Gradio interactive classifier for ischaemic vs. haemorrhagic stroke with Grad-CAM | App Built |
| PSO-Optimized ResU-Net for Brain Tumour | Bhavani J (M.Sc. Data Science) | Dr. Naresh Kumar | BraTS 2019 benchmark: Dice 0.871, IoU 0.798; Streamlit patient report generator | App Built |
| Range-Based Classifier for Tabular Data | Sakthi Aakarshan R (M.Sc. DS) | Dr. B. Santhi | Interpretable rule-based classifier matching KNN and Random Forest on cancer data | Validation |