Working as a NSF Computing Innovation Postdoctoral Research Fellow in the Causal AI lab of Dr. Elias Bareinboim (https://causalai.net).Connecting causal inference and computational neuroscience.- causal discovery of high-dimensional multivariate time series with latent confounding - fundamental causal discovery with observational and experimental data in a multiple environmental setting- statistical modeling and information theory

Core-developer helping maintain and develop an open-source codebase with 59k stars and millions of downloads. I perform the following tasks:- review pull requests from developers- implement features, such as missing-value support for extremely randomized forests, or metadata routing in sklearn meta-estimators- contribute to technical discussions on API design, bugs, and Cython/Python technicalities

AAMPLIFY is an education and charitable nonprofit dedicated to promoting Asian Pacific American leaders in public service. By empowering and mentoring low-income, first-generation high school students, AAMPLIFY seeks to invest in the next generation of leaders to find solutions to issues that affect our communities. Together, our voices are amplifed. To learn more or get involved, please visit: aamplify.us.org

- Led research & development of a causal machine learning model (applied to 100M+ samples) to dynamically match users with promotional campaigns demonstrating a potential 3-8% increase in profit margins for USA Eats platform.- Developed solution to enable Python3.8+ in PySpark and SparkMagic Jupyter notebooks, enabling users to upgrade and reduce technical debt in data science workflows with Python, Hive and Hadoop

Wrote compile-time code using C++ template meta-programming for sparse N-dimensional arrays: https://github.com/pydata/sparse- implement custom iterators- implement template-recursion for compile-time checks- implement template API for code-generation of efficient sparse tensor operations- integrate C++ API that adheres to the functionality of http://tensor-compiler.org

Research, product development, customer discovery, and software engineering (https://neurologicsolutions.net).Raised over $600K to-date (NSF SBIR Phase I, Mayland Innovation Initiative, JHTV Pitch Competition, NSF SBIR Phase I).Filed provisional patents and full patents in the USA, European and Japan markets through collaboration with Johns Hopkins Technology Ventures (JHTV).Led 510k FDA approval process with a team of 5 engineers, consultants and advisors involving risk analysis, software requirements, design specifications, and user-testing (unit testing, continuous integration, and software documentation). Led development of MVP that could be used by clinicians via a web-application. Backend consisted of AWS, Kubernetes, Docker, Python based REST API and a Flux-enabled continuous delivery solution.

My research is focused on studying epilepsy and understanding the network characteristics of the epileptogenic zone (EZ; the origin of seizures). This involves analyzing neuroimaging data, such as T1 MRI, CT and intracranial EEG signals from epileptic patients. I work with different clinical teams across the USA. I aggregate and organize electrophysiological data of epileptic patients from 4 different hospital centers by coordinating with neurosurgeons, epileptologists, and fellows in setting up a HIPPA-compliant sFTP server. I work on both data analysis, as well as computational models of epilepsy at a population dynamics level; I work on end-end data analysis from aggregation, wrangling, analysis to visualization. My responsibilities and work have included:• Developing mathematical models of epilepsy using linear systems, matrix perturbation theory, graph theory and probability• Analyzing and maintaining (with HIPAA sFTP) terabyte’s of multivariate time series, categorical and imaging data from four epilepsy centers around the USA• Engineering an data analysis pipeline using parallel programming to compute MVAR and perturbation models to predict the seizure onset region within the brain (Python, MATLAB, SLURM, GNU, multithreading)• Developing algorithms for electrode noise filtering, frequency analysis and Bayesian analysis of data (Python, PyMC3)Computing Familiarity:- Familiar with SLURM, PBS, GNU Parallel and running high-performance-computing parallel jobs using CPU and GPU clusters- Familiar with bash and Linux- I play around with Cython and IPython magic commands to optimize all scientific computing code from Python and profiling codeSummary:I utilize high-performance computing to run Matlab and Python code analyzing large amounts of electrophysiological human data.

1st Rotation (Neuromedical Control Systems Lab: http://sarmalab.icm.jhu.edu/):Data collected in collaboration with the Johns Hopkins Hospital from epileptic patients.Using network based analysis and machine learning algorithms, I am exploring the notion of a pre-seizure state. I analyze the dynamics of the brain network over time using various signal processing techniques and concepts. I explored results using, modularity optimization, spectral factorization, k-means, eigenvector centrality and granger causality.2nd Rotation (Computational Biophotonics Lab):Colorectal cancer (CRC) is the third leading cause of cancer death in the world. Approximately 1/60 deaths in developing countries is from CRC. There is recent evidence that screening colonoscopies only reduce chance of death from CRC by ~29-40%. Missed lesions occur due to poor lesion contrast and time constraints.Using photometric stereo, we can reconstruct colonoscopy videos into topographical surface maps to enhance surface information to allow GIs to better determine a lesion. I will explore the use of a Phong reflectance model in photometric stereo endoscopy and also using machine learning to build a computer-assisted system for identifying lesions.3rd Rotation (Functional and Restorative Neurosurgery Unit @ National Institute of Health):https://neuroscience.nih.gov/ninds/zaghloul/Studying memory using data science because it's super cool.Languages: Python, MATLABTools: IPython

Implement statistical connectivity package for MEG, EEG and iEEG data:https://github.com/mne-tools/mne-connectivity/Python API, Continuous integration, technical documentation and numerical computing

Funded by the Whitaker, Chateaubriand, and NSF-GRFP fellowships, I came to work within the Theoretical Neurosciences Group (TNG) at Marseille University for one year. TNG wants to uncover the mechanisms underlying the spatiotemporal organization of large-scale brain networks. I am involved in various projects, which can be summarized in two research directions: 1) Using whole-brain models and state-of-the-art data analytics to study how different parameter settings within the brain can affect algorithm analysis and 2) generating realistic and diverse seizure data using nonlinear models to train deep neural networks to perform detection, localization and prediction on real seizure data, thus overcoming "data limitations" currently required for deep learning. My responsibilities include the following:• Using Bash, MATLAB, Python and C++ to analyze and preprocess > 5TB of multi-modality brain imaging data for localizing electrode contacts, analyzing region activity and visualizing data-embedded brains• Engineering a transfer learning pipeline using nonlinear generative modeling, linear stability analysis and artificial neural networks (FNN, CNN, RNNs)• Developing nonlinear dynamical stochastic mass models to optimize algorithm parameters that have shown significant results (>95% accuracy) in identifying the seizure onset zone Advisors: Dr. Viktor Jirsa, Dr. Sridevi Sarma

I was the co-founder of a startup while I was in my last few years of undergraduate university. We focused on data analytics for Parkinson's disease and creating toolsets for clinicians to track the disease in a more quantitative fashion. I worked with a colleague and also mentors through San Diego’s entrepreneurship ecosystem. Partner: Neil Gandhi (future M.D.)

• Assist 100+ students in learning basic data structures in Java, C and C++• Grade exams and assist professor in communicating fundamental concepts in computer science