Home > Center for Artificial Intelligence Research (CAIR) > Our Projects

Our Projects

  • CAIR

Engineering Intelligence for The Real World

At the Center for AI Research (CAIR), we are cultivating a vibrant and forward-looking research ecosystem focused on solving real-world challenges through intelligent, scalable technologies. Our work spans key frontiers of innovation—Edge AI, High-Performance Computing (HPC), Data Science, and Computational Medicine—driving solutions that are not only fast and reliable, but also future-ready.

CAIR bridges the gap between foundational research and applied innovation. By fostering strong collaborations with industry leaders, academic institutions, and technology partners, we ensure that our breakthroughs go beyond the lab to create measurable impact. From accelerating healthcare insights to enabling intelligent infrastructure, CAIR is where engineering meets intelligence, and ideas evolve into real-world solutions.



Engineering Intelligence
  • EDGE AI PROJECTS

Smarter Models. Lighter Systems.

At CAIR, our Edge AI research is focused on bringing real-time intelligence to the edge—where devices operate with constrained computing resources, yet demand high responsiveness and reliability. We develop lightweight, high-performance AI models that are optimized for speed, scalability, and accuracy.

From healthcare diagnostics to smart manufacturing, our projects apply cutting-edge techniques such as model compression, quantization, and task-specific adaptation to solve real-world problems—efficiently and at scale.

  • Redundancy & Quantization-Based Optimization of DNNs
  • This project addresses pressing challenges in healthcare AI—such as medical image segmentation, disease detection, and risk prediction—by making deep neural networks (DNNs) lighter and faster. Our team explores redundancy reduction and quantization techniques to compress models without compromising performance. These optimized models are tailored for multi-modal health data analysis and are deployable on edge devices, enabling cost-effective and energy-efficient medical diagnostics at the point of care.
  • Filter Pruning for Efficient Video Classification
  • In video classification tasks—used in surveillance, robotics, and autonomous navigation—latency and compute load are major constraints. This project introduces a novel Stripe-Wise Pruning (SWP) method, enhanced by the SparPen optimizer, to selectively prune filters in convolutional layers. The result: lighter, faster models with significantly reduced memory and computation requirements—while retaining high accuracy and responsiveness for real-time deployment.
  • Domain Adaptation and Fine-Tuning for GenAI Models
  • To make Generative AI (GenAI) models more task-aware and efficient, this project focuses on domain adaptation and fine-tuning of large pre-trained models. By customizing these models for specific applications—such as text-to-image generation and multimodal learning—we reduce training time and data demands while improving output relevance and accuracy. This research paves the way for context-sensitive, low-footprint GenAI that can run efficiently on local systems or edge platforms.
  • Industry Collaboration: Intelligent Assembly Line at Spicer-DANA
  • In a strategic partnership with Spicer-DANA, CAIR developed an AI-powered quality validation system for automotive assembly lines. The goal: reduce inspection errors and productivity losses caused by manual checks. The solution: a lightweight computer vision model, optimized for deployment on embedded systems, and seamlessly integrated into the assembly workflow. This real-world deployment not only reduced costs and improved consistency, but also demonstrated the transformative potential of edge-deployed AI in industrial settings.
  • HPC PROJECTS

Compute Power. Scientific Precision.

At the Center for AI Research (CAIR), our High-Performance Computing (HPC) initiatives tackle large-scale scientific, engineering, and real-time challenges by combining deep learning with cutting-edge hardware acceleration. From 3D rendering to hyperspectral analysis and quantum modeling, our projects harness the full power of parallel computation to drive breakthroughs across disciplines.

Our HPC research is closely aligned with national missions such as the National Supercomputing Mission (NSM) and the National Quantum Mission, and is powered by collaborations with institutions including Karnataka Medical College, SDM University, and IIIT Dharwad.

  • Quantum Software Development Lifecycle for Diabetic Retinopathy
    Sponsor: MeitY-QCAL | Duration: 2023–2025 | Grant: $19,860
  • This groundbreaking project focuses on developing a Quantum Software Development Lifecycle (QSDLC) tailored for diabetic retinopathy classification. Led by Dr. Satyadhyan Chickerur (PI) and Dr. K. M. M. Rajeshekariah (Co-PI), the research aims to create a structured framework for translating classical AI models into quantum-compatible applications. The goal is to enhance diagnostic precision while bridging classical and quantum computing for next-generation medical diagnostics.
  • Automated Grading of Diabetic Retinopathy at PHCs
    Sponsor: DST–NSM | Duration: 2021–2023 | Grant: ₹19.886 lakhs
  • Led by Dr. Satyadhyan Chickerur, with Dr. Mahesh Patil and Dr. Shantala Giraddi, this project developed a distributed deep learning framework to automate the grading of diabetic retinopathy. By transferring optimal hyperparameters (OHPs) from the global EyePACS dataset to India’s IDRiD dataset, the team eliminated the need for costly retraining—enabling scalable, efficient deployment at Primary Healthcare Centers (PHCs). The outcome: faster, more accurate diagnostics in rural and resource-limited settings.
  • Blood Flow Analysis using Deep Learning and CFD
  • This project combines MRI-based image processing with Computational Fluid Dynamics (CFD) to model hemodynamic behavior for cardiovascular diagnostics. Deep learning algorithms extract vascular flow data, which is simulated in CFD environments to provide detailed, patient-specific insights into blood flow patterns—paving the way for early diagnosis and treatment planning.
  • Power-Efficient 3D Rendering on Multi-Core Processors
  • This research explores energy optimization in GPU-based 3D rendering for applications in VR, gaming, and simulation. By analyzing parameters like shader complexity, scene load, and pipeline bottlenecks, the team identifies strategies to reduce power consumption without compromising performance—making rendering engines more sustainable and scalable across devices.
  • Deforestation Monitoring using Hyperspectral Imagery
  • Using semi-supervised deep learning and semantic segmentation, this project detects and tracks deforestation patterns with high spatial and spectral resolution. By analyzing hyperspectral satellite data, the system supports fine-grained environmental monitoring, aiding in conservation policy, ecological planning, and early intervention in deforestation-prone regions.
  • DATA SCIENCE PROJECTS

Meaningful Data. Measurable Impact.

At CAIR, our data science research harnesses the potential of artificial intelligence to build clinically relevant tools that improve diagnostics, monitoring, and care delivery. From cardiovascular imaging and neonatal wellness to assistive technologies for visual impairment, our projects demonstrate the life-changing impact of data-driven models in healthcare.

These initiatives are developed in close collaboration with KLE Kore Hospital, JNMC KAHER, and Hubli Scan Center, ensuring that our research is both medically grounded and technologically advanced.

  • Coronary Artery Plaque Detection from 3D CCTA Images
  • Partner: KLE Kore Hospital, Belagavi
    This project automates the segmentation and analysis of coronary arteries from 3D Coronary Computed Tomography Angiography (CCTA) scans. By accurately identifying plaque deposits and assessing vessel structures, the AI model assists clinicians in detecting early-stage cardiovascular disease, improving treatment planning, and reducing diagnostic uncertainty.
  • Kangaroo Mother Care (KMC) Monitoring – KANGA Suraksha
  • Partner: JNMC KAHER
    This project introduces KANGA Suraksha, a real-time monitoring device that tracks compliance with Kangaroo Mother Care (KMC) protocols—critical for the survival of preterm and low birthweight infants. By combining sensor-based tracking with data analytics, this solution reinforces maternal-infant bonding practices and advances neonatal care innovation.
  • Cortical Visual Impairment (CVI) Support Application
  • Partner: KLE Kore Hospital
    Designed as a therapy-aligned assistive tool, this application supports individuals—particularly children—with Cortical Visual Impairment (CVI). The platform promotes visual engagement, cognitive development, and independence, while offering structured support for therapists, educators, and families involved in CVI care.
  • Metal Artifact Reduction (MAR) in CT/MRI Scans
  • Partner: Hubli Scan Center
    Metallic implants often create distortions in medical imaging, complicating diagnosis. This project develops a Metal Artifact Reduction (MAR) system that refines CT and MRI scans to eliminate visual interference caused by prosthetics or surgical hardware—enhancing diagnostic clarity and supporting more precise treatment decisions.
  • Low-Dose CT Image Enhancement using FreeSeed Model
  • To address the trade-off between radiation exposure and diagnostic detail, this project leverages the FreeSeed model—a dual-domain AI network that blends frequency-aware and sinogram-based learning. It enhances low-dose, sparse-view CT images by restoring missing anatomical structures, supporting safer imaging without compromising accuracy.
  • PREDICTIVE ANALYTICS

Forecasting with Intelligence

At CAIR, we bridge machine learning and domain-specific insights to develop predictive models that guide real-world decisions. Our work in agriculture and environmental sciences empowers stakeholders—researchers, farmers, and policymakers—to anticipate trends, mitigate risks, and act proactively.

  • Late Leaf Spot Detection and Groundnut Pod Quality Assessment
  • Partner: University of Agricultural Sciences, Dharwad
    This project develops a deep learning model to detect Late Leaf Spot (LLS)—a fungal disease in groundnut crops—and assess its impact on pod quality. By correlating disease severity with yield outcomes, the model provides actionable insights for early intervention, crop health monitoring, and precision agriculture practices.
  • RESEARCH FOCUS AREAS

Domains of Impact

01

Edge AI

02

High Performance Computing

03

Data Science

04

Cloud Computing