About Levitating Labs

Welcome to Levitating Labs — an open research notebook dedicated to AI in Drug Discovery.

Behind the Lab

Hiren Limbad

Final Year B.E. in Computer Engineering

Building machine learning and graph neural network architectures for biomedical discovery. Explore my projects on my Portfolio Website.

Why an Open Research Notebook?

• ⚡ Rapid Knowledge Sharing: Publish empirical benchmarks, ablation studies, and negative results as soon as experiments finish.
• 🔬 Complete Transparency: Release unvarnished metrics, code artifacts, and datasets without waiting for publication pipelines.
• 🤝 Immediate Peer Feedback: Enable the research community to inspect methodologies, reproduce results, and collaborate open-source.

Lab Rules & Core Principles

We adhere strictly to empirical rigor, scientific transparency, and benchmark integrity:

1. No Shortcut Evaluation
   Evaluation data is strictly leakage-free: conformer generation, molecular scaffold, and train/test splits are kept physically disjoint.

2. Standardized Baselines
   Every experiment is benchmarked against established baselines—ensuring honest, transparent reporting even when there is no performance gain.

3. First-Principles Research
   We prioritize solving fundamental scientific bottlenecks over chasing minor hyperparameter gains on saturated benchmarks.

Research Focus Areas

• 🧬 Graph Neural Networks & Geometric DL for Molecules
• 💊 Conformer Generation & Structural Docking
• 🎯 Scaffold Generalization & Out-of-Distribution Evaluation