Collins Lab

Collins Lab

• Designed and implemented a mobile application, using Swift and React Native, that interacts with hardware technology to facilitate impulse buying for a pre-seed stealth fashion startup seeking to transform impulse fashion consumption.• Experienced startup environment as one of first non-founding members and learned about venture capital funding and other aspects of startups.

• Center for Molecular and Cellular Neuroscience – Gradinaru Lab• Advisors: Prof. Viviana Gradinaru, Dr. Min Jee Jang, Dr. Anat Kahan, Dr. David Brown• Worked on development of a single-cell sequencing and machine learning pipeline to characterize adeno-associated viral (AAV) tropisms at high resolution.• Trained imbalanced classification models to pioneer a "virtual knockout" methodology for identification of genes whose expression facilitates or inhibits AAV transduction.• Worked on development of an experimental and computational pipeline leveraging fluorescence in situ hybridization and spatial transcriptomics to characterize AAV tropisms.• Developed machine learning models for classification of VIP neurons in the suprachiasmatic nucleus and prediction of neuron type based on calcium imaging.• Published as co-author in Nature Biotechnology and Frontiers of Immunology.

• Division of Biology and Biological Engineering – Lester Research Group• Advisors: Prof. Henry Lester, Dr. Anand Muthusamy• Worked on development of Inside-Out, a suite of web apps for simulation of drug concentrations and receptor activity during ingestion and elimination of commonly abused drugs.• Analyzed effects of nicotine dose and cytochrome P450 2A6 polymorphisms on activation and chaperoning pathways of nicotine addiction.• Engineered genetically encodable opioid biosensors via directed evolution.• Pioneered computational methods for automating detection of ethologically relevant behavioral responses to opioids in mice from markerless pose estimation readout.

• Held weekly office hours for an applied mathematics course, IDS/ACM/CS 157: Statistical Inference.• Held review sessions for the midterm and final exams.• Assigned problems each week to team of teaching assistants for grading.• Compiled all students' midterm and final grades to record course progress.

• Tutored classmates in various Computer Science (CS), Applied and Computational Mathematics (ACM), Bioengineering (BE), Computation and Neural Systems (CNS), Economics (Ec), Mathematics (Ma), Physics (Ph), and Biology (Bi) courses.

• Held weekly office hours and graded problem sets for a graduate-level applied mathematics course, ACM/IDS 216: Markov Chains, Discrete Stochastic Processes and Applications.• Compiled all students' final grades to record course progress.

• Decision, Optimization and Learning at Caltech (DOLCIT) – Yue Group• Advisors: Prof. Yisong Yue, Dr. Jialin Song, Dr. Yury Tokpanov• Integrated deep kernel learning into multi-fidelity Bayesian Optimization algorithms to improve model performance.• Applied resulting optimization methods to analyze astronomy and nanophotonics datasets.

• Founded a startup aiming to use analytics to provide personalized restaurant recommendations.• Designed and implemented a mobile application, using Swift, that learns user dining preferences and presents recommendations.• Awarded the Mayleben Venture Shaping grant by the Zell Lurie Institute of the University of Michigan’s Ross School of Business.

• Designed and implemented register allocation algorithm to optimize translation of deep learning models from open-source deep learning frameworks into MATLAB.• Constructed framework to adapt computer vision models for application in highway lane identification and following.

• Trained machine learning models to predict editing efficiency of CRISPR systems based on secondary structures of guide sequences.

• Department of Psychiatry – Lifshitz Group• Advisor: Prof. Michael Lifshitz• Studied the structural and functional correlates of imaginative suggestibility via analysis of behavioral and fMRI data.• Conducted network analysis of psychological traits in neurophenomenology survey data.

• Trained deep learning classifiers on chest CT scans for COVID-19 detection based on various radiological features, including ground glass opacity and crazy paving patterns.

• Performed statistical analysis on order imbalance and price data of 500 individual tickers and built trading models and devised quantitative trading strategies.• Implemented simulations to test and evaluate trading strategies.

• Conducted industry research in the computational drug design/discovery and neuropharmacology space and sourced early-stage deals and startups.• Sought out early-stage emerging technology with significant industry-affecting potential.• Riot Ventures is an LA-based, globally focused fund investing in early stage deep tech companies.

• Department of Anthropology – Luhrmann Research Group• Advisors: Prof. Tanya Luhrmann, Prof. Michael Lifshitz• Worked on design and development of a neuroimaging paradigm to investigate brain mechanisms involved in auditory verbal hallucinations in hallucination-prone individuals.• Built psychological tasks integrated into an fMRI protocol to simultaneously evaluate behavioral and brain activity data.• Constructed pipeline to process and analyze fMRI and behavioral data.

• Worked as a member of the Information Technology Team.• Designed and implemented software to construct and visualize network effects and interactions among attributes of securities (specifically bonds).• Trained clustering models on fixed income investment data to classify securities based on interactive effects and improve portfolio management strategies.• Trained anomaly detection models to identify faulty entries in time series portfolio data.

• Trained supervised deep learning models for classification of images taken by the HiRISE instrument of the Mars Reconnaissance Orbiter.• Analyzed classification performance of various model architectures.

• Worked as a member of the Network Engineering and Platform Team.• Analyzed network flow time series data to optimize allocation of network traffic across available routes.• Trained stateful Long Short-Term Memory (LSTM) Recurrent Neural Networks to forecast network traffic and flow, using TensorFlow and Keras.• Applied model to forecast network activity of a major company client.

• Center for Genomics and Personalized Medicine – Urban Lab• Advisors: Prof. Alexander Urban, Dr. Xiaowei Zhu• Worked on training, implementation, and packaging of RetroSom, a tool for classification of transposons via a transfer learning model.• Trained an imbalanced classifier of true and false mobile element insertions (MEIs) in Python that achieved 0.96 Area Under Precision-Recall Curve (AUPRC).• Acknowledged in Nature Neuroscience paper.

• Chan Zuckerberg Biohub – Elias Lab• Advisors: Prof. Joshua Elias, Dr. Lichao Zhang• Analyzed and modeled the effects of noise and interference on the detection of biomarkers in tandem mass tags mass spectrometry data.• Trained neural networks, in Java and Mathematica, to model and filter out noise in mass spectrometry data.