I'm a graduate student in Artificial Intelligence at Boston University, with a deep interest in building AI systems that can reason, evolve, and interact with biological data. My work bridges neuroscience, machine learning, and symbolic abstraction — with the goal of building more adaptive and insightful models of the world.
Darwin is a self-evolving research architecture that simulates distributed scientific reasoning. It combines:
- Agent-based theory evolution
- Bayesian optimization and causal attribution
- Symbolic motif reuse and contradiction compression
- Interfaces with both simulated and real neural feedback systems
An advanced medical image segmentation pipeline using:
- Transformer-enhanced 2.5D nnUNet
- Conditional diffusion models (Med-DDPM) for rare tumor synthesis
- Cross-modality fusion of T1 and T2 MRIs This project aims to improve clinical tracking of vestibular schwannomas through robust, size-aware automation.
HEG is a scalable system for shortest path queries and abstract reasoning over knowledge graphs. It integrates:
- Community-level graph summarization
- Per-community node2vec embeddings
- Bitmask indexing for fast path rejection
- Hybrid symbolic and learned routing strategies
- 📫 Email: eliconlin@gmail.com
- 🔗 LinkedIn: linkedin.com/in/eliconlin
“Build models that don’t just predict — but evolve, question, and explain.”