Advanced Internship Program on Generative AI

Phase II Track Expansion

Following the successful deployment and evaluation of our Phase I Summer Research Internship on Vision Tasks (YOLO), iHub-Data at IIIT Hyderabad is expanding the track to focus on modern Generative AI execution layers.

This 2-to-3-month advanced track transitions selected high-performing interns from standard deep learning classification/localization to fully customizable generative mechanics. Utilizing locally provisioned instances of ComfyUI alongside linked system node variations, interns will map complex vision logic directly onto modern generative layers.

Core ComfyUI Exploration Pillars

Unlike standard black-box generative applications, ComfyUI offers granular structural manipulation over the complete lifecycle of a model execution. Interns will dive deep into:

• Graph-Based Production Logic: Constructing modular mathematical workflows by separating text weights, noise conditioning nodes, scheduling steps, and latent space representations.
• Controlled Architecture Synthesis: Using state modifiers like ControlNet and T2I-Adapters to restrict generated boundaries based on edge contours, custom depth metrics, or camera layouts.
• YOLO-to-Generative Bridges: Mapping your Phase I detection output bounding boxes to automatically mask, isolate, and procedurally generate context-aware transformations on specific image targets.