Worked with University of Birmingham researchers on machine-learning approaches for digital chemistry. The work was conceptually related to the later AI drug-discovery dissertation because it explored how ML could represent and reason about molecular structures, while the internship leaned further toward computer vision: molecular skeleton representations, image-like structural inputs, autoencoder-style experimentation and visual validation of learned representations. The project involved testing architectures, communicating results to technical domain experts and producing recommendations for future research.
Research focus
Computer vision for molecular structure understanding
This was an earlier digital-chemistry research project in the same broad family as the dissertation work. Instead of building a full generative drug-discovery pipeline, it concentrated on how visual molecular skeleton representations could be processed with machine-learning architectures.
- Molecular skeleton representations
- Computer-vision-led architecture experiments
- Autoencoder and representation-learning ideas
Digital chemistry
Domain
Computer vision
Focus
Research report
Output
Key details
- Explored ML architectures for molecular skeleton representations with a stronger computer-vision emphasis than the later final-year dissertation.
- Worked with image-like and structure-focused molecular inputs to test how visual representations could support digital chemistry workflows.
- Experimented with autoencoder-style approaches and representation learning for compact molecular structure understanding.
- Reported findings to technical experts and recommended further experimentation for future digital chemistry research.