Abdul Basit

Modeled complex network topologies and developed performance-optimized GNNs for anomaly detection. Adapted ANEMONE into an edge-aware GATv2 self-supervised framework. Designed a physics-informed synthetic anomaly generator. Built end-to-end ML pipelines with distributed training on HPC clusters using Slurm.

Built an agentic DC-DC converter design platform on a strict deterministic/LLM separation: a LangGraph ReAct agent orchestrates. Integrated SPICE transients (PySpice), Type-III control loops, 3D CHT thermal solves (OpenFOAM), and NSGA-II optimization. Validated against real TI evaluation hardware, with Dockerized environment, Streamlit dashboard, and a 200+-test suite.

Multimodal Text-Guided Video Segment Search

Conducted a comprehensive study on state-of-the-art techniques for text-guided temporal video retrieval. Focused on retrieving specific moments based on natural language queries by studying uniform, adaptive, and dynamic sampling strategies.

Multimodal Vision-Language Models for Remote Sensing

Evaluated Qwen2.5-VL and InternVL3.5 on a custom 20,000-example benchmark spanning RGB, SAR, and multispectral data. Addressed cross-modal fusion bottlenecks via weighted fine-tuning and test-time ablation studies, producing a novel bias taxonomy for robust Earth observation AI.

3-D Reconstruction of Scenes from Images

Analyzed state-of-the-art ML techniques for 3D reconstruction from 2D images, including NeRF, Instant NeRF, PixelNeRF, Gaussian Splatting, and Neuralangelo.

Learning Arithmetic Operations Using Small-LLM With Reinforcement Learning

Developed a lightweight LLM specialized in arithmetic reasoning using a two-stage pipeline: Supervised Fine-Tuning on a curriculum-based dataset, followed by Reinforcement Learning via Expert Iteration (Rejection Sampling).

Event Cameras

Investigated operating principles and algorithmic applications of Event-based Cameras (Dynamic Vision Sensors), including data visualization, spatiotemporal analysis, and motion estimation.

Extended Reality – Towers of Hanoi

Designed an immersive XR application using WebXR and Three.js, ensuring cross-platform compatibility on devices such as Oculus Quest.

Continual Learning for Multi-Image Classification

Evaluated strategies for mitigating Catastrophic Forgetting using the Split-MNIST benchmark, including Elastic Weight Consolidation (EWC) and Experience Replay.

GAN vs VAE

Comparative analysis of GANs and VAEs: sample quality, training stability, latent space representation, and generalization across different generative learning paradigms.

Style Transfer & GAN-Based Data Augmentation

Built custom CNNs, developed Conditional GANs with progressive resolution growth, and implemented CycleGAN for unpaired image-to-image translation.

AI-Based Plant Monitoring System (Funded)

Deployed 5 CNN architectures (ResNet50, VGG16/19, MobileNet V1/V2) on Raspberry Pi using TF Lite for real-time disease detection. Integrated with a Cable-Driven Parallel Robot.

Designed PLC code for process automation, resolved PLC system issues, and upgraded HMI code for plant monitoring.