Loading signal
Preparing the next layer of the atlas.
CORE
Education
M.E.
Bioengineering
UC San Diego
2021
B.S.
Bioengineering: Biosystems
UC San Diego
2020
Stack
languages
ml/dl
architectures
data science
cloud
aws services
frameworks
neuro tools
data formats
signal processing
analysis
visualization
devops
tools
Domains
Affiliations
Panoptic Bio
Founding Applied AI Scientist
2026
Stealth NeuroAI Startup
Founding Engineer
2025
DataJoint
Neuroscience Data Engineer II
2022–2024
NEATLABs, UCSD
Lead Lab Programmer
2017–2022
Dolby Laboratories
Engineering Intern
2014–2016
Analysis Methods
calcium imaging
ROI detection and signal extraction
Constrained NMF for calcium imaging
Robust cell extraction
Non-rigid motion correction
pose & behavior
Markerless pose estimation
Orofacial and behavioral tracking
Population activity from GCaMP signals
electrophysiology
Unified spike sorting API
Template-matching spike sorter
Allen Institute pipeline
High-density probe preprocessing
behavioral modeling
Reinforcement learning models
Belief updating under uncertainty
Temporal preference modeling
Impulse control metrics
neural time-series
Power spectra and band power via FFT
Time-frequency decomposition
Instantaneous amplitude and phase
Phase locking and phase-amplitude coupling
Frequency-resolved coupling between sites
Connectivity across all bands; beta/delta significant in paper
Pairwise neural synchrony
Directed functional connectivity
Trial-locked responses and spectral perturbation
brain encoding
Self-supervised ViT visual representations
Motion features for video encoding
Regularized feature-to-BOLD readout
Region-specific reliability scaling
Falsification-first evaluation
mechanistic interpretability
Hooked-transformer activation introspection
Circuit discovery via path interventions
Causal localization of computation
Automated circuit-graph search
Behavior under edited activations
Decoding features from representations
Head-level pattern inspection
Directed edits and manifold analysis
foundation models
Masked and contrastive objectives
Attention over extended neural context
Linear-time sequence modeling
Cross-attention over heterogeneous inputs
Latent compression of neural signals
Parameter-efficient fine-tuning
Sharded large-model training
Stitching subjects and sessions
multimodal systems
Video, audio, text, and neural tokenizers
Joint attention across streams
CLIP-style representation binding
Linear / ridge encoding heads
Robustness to missing inputs
Harmonized tokens across sensors
data infrastructure
Workflow and provenance modeling
Neurodata Without Borders format
Chunked array storage
AWS, parallel pipelines
Deployed Systems
Deployed cloud-based DeepLabCut, Facemap, and calcium imaging pipelines for the Lu Lab at UCSD.
Large-scale electrophysiology processing with same-day data upload and reproducible pipeline execution.
DataJoint workflows for calcium imaging, electrophysiology, fiber photometry, and behavioral tracking.
Applied AI systems for clinical trial intelligence: data pipelines, model evaluation, and decision-support tooling.
Links