ML · Scientific Computing · Reproducible Simulation Systems
Building continuous-time dynamical models, simulation validation infrastructure, and graph-based biological systems
 Google Summer of Code 2026 ✅ Accepted |
× |
 INCF Mentoring Organization |
× |
 HNN-Core Project Repository |
Project: Refactoring Synaptic Behavior in HNN-Core
Selected as a GSoC 2026 contributor under INCF to improve biological realism in HNN-Core — placing synapses only at active connection sites, making synapse position along sections configurable, recording currents only from active synapses, and adding seeded biological variability to synaptic weights and delays.
Primary Mentor: Katharina Duecker · Mentors: Nicholas Tolley, Austin Soplata, Anna Cattani
Working at the intersection of ML and computational neuroscience — where correctness is verifiable and reproducibility is enforced by architecture.
GSoC 2026 contributor to HNN-Core (INCF), shipping production-quality changes to NEURON-based synaptic simulation internals.
Focus: Neural ODEs · Modular simulation pipelines · Structured validation frameworks for scientific Python
Neural ODEs · Dynamical Systems · Simulation Pipeline Design
Numerical Validation · Solver Benchmarking · Graph Neural Networks
Biological System Modeling · ONNX Deployment · Scientific Python
NEURON Simulation · Synaptic Modeling · Computational Neuroscience
| Project | Stack | What it does |
|---|---|---|
| HNN Simulation Validation | Python · NumPy · SciPy · pytest | Modular validation pipeline for neural simulation outputs — peak latency checks, waveform integrity, composable validator modules per drive type |
| Neural ODE Cognitive Simulator | PyTorch · torchdiffeq · FastAPI | Continuous-time model dS/dt = f(S,U,θ) of EEG-derived cognitive states using adaptive solvers (dopri5, RK45) |
| Metabolic Pathway GNN | PyG · GATv2 · FastAPI · React | Glycolysis / TCA / PPP as bipartite graphs; GATv2 + FiLM for flux prediction, validated against literature ranges |
| AirAware — Browser AQI | ONNX Runtime · React · Tailwind | AQI model exported to ONNX, deployed client-side; <0.1% deviation vs Python outputs confirmed |
