- Co-founded an AI company that uses ML to identify and fix vulnerabilities in smart contracts, raising $1.7M from investors like Alchemy and Symbolic Capital with advisors/angels from Meta, Apple, and Ledger- Trained a lightweight graph neural network (GNN) with NetworkX and PyTorch Geometric to classify cross-contract and cross-function reentrancy from control flow graphs (CFG)- Spearheaded an effort to use SFT and RLHF to finetune and steer LLMs (Llama2-70B and Mistral-Medium) into identifying vulnerable code that evaded all traditional static analyzers like Slither- Augmented a fuzzer with a small Transformer fitted with a custom vocabulary and tokenizer to filter transactions based on the likelihood that they broke predefined invariants created in Echidna- Established a robust hosting and infrastructure environment using AWS SageMaker, Docker, and Modal for efficient model training, testing, and deployment, streamlined with Terraform for seamless automation and scalability

- Collaborated with postdocs to create a shallow U-Net in TensorFlow to segment neurons in 2-photon calcium imaging videos called SUNS- Created an active learning pipeline that selectively labels neurons based on their frequency and SNR in successive SUNS runs, which reduces the need for labeled neurons to reach SOTA by 50x- Developed sets of Bash scripts for automating jobs across GPU clusters and monitored active learning jobs using custom Python scripts to probe aggregated results

- Implemented an approach in PyTorch to estimate training data influence by tracing gradient descent with Weights and Biases logs and Torch Studio- Extended approach to object detection for detection of false negatives in human-labeled datasets; treated mislabeled instances as its own class based on mixed confidences from its original labeled classes- Detected over 50% of false negatives in self-driving datasets and reported them to human labelers as a form of feedback; deployed feedback model on AWS SageMaker

- Led a team of 3D engineers and ML research scientists in leveraging domain-randomized synthetic data to train product recognition algorithms, utilizing Unity for simulation and PyTorch for model development, hosted on Azure Cloud for scalable processing- Developed a product auto-labeling pipeline that utilizes instance segmentations from simulation data, streamlining the labeling process and enhancing model training efficiency- Created photorealistic data by using cycle-consistent GANs to transform unpaired real and simulation images, closing the domain gap improved model generalization to in-store data- Implemented a hybrid training approach using synthetic data to pretrain YOLOv5 and a small set of real examples for Supervised Fine-Tuning (SFT), significantly reducing store deployment time from 2 weeks to 2 days and boosting new SKU detection by 80%

- Conducted experiments to increase the NGFR+ percentage (editing rate) for FOXP3 gene editing in hematopoietic stem progenitor cells (HSPCs) at the Bacchetta Lab- Utilized a design of experiments approach to optimize CRISPR-Cas9 editing of HSPCs, resulting in a published paper in Cytotherapy Journal- Performed cost analysis to demonstrate a reduction in the cost of reagents for gene editing experiments- Designed and executed experiments using varying amounts of CRISPR-Cas9 reagents, and analyzed flow cytometry data to assess editing efficiency and cell viability

- Collaborated with KORA (Kinetic Operating Room Assistant) to develop a camera-based vision system as a foundation for machine learning solutions in the operating room, focusing on autonomous lighting and camera systems for image acquisition and touchless operating area illumination- Utilized Python scripts and BLE (Bluetooth Low Energy) technology to fetch real-time 3D coordinates of the doctor's wrists, enabling precise control of the operating area's lighting- Developed a BLE-based system to dynamically adjust the light source, ensuring optimal illumination over the designated coordinates during surgical procedures