Improved and productionized data pipelines to train variational auto encoders for anomaly detection on over 2000 endpoint behaviorsCreated and maintained a neo4j graph database to visualize and track customer networking behavior and its graph based propertiesEmployed RAG and prompt engineering for efficient one shot extraction of salient information from knowledge graphsCreated and improved agentic workflows to 3x efficiency of cybersecurity researchCreated pipelines to A/B test LLM fine-turning deployments and utilized Langfuse to monitor cost and performance

Utilized LSTMs and voice analytics to extract emotional undertones from human speechDeployed production ML model on AWS API gateway and Lambda

Created pipeline to parse and transfer 2TB of SEC EDGAR filings to Open Search cluster for ease of querying during model trainingTrained 50k feature tokenizer and pretrained BERT Base model on EDGAR dataTrained and improved Named Entity Recognition model to extract causes and effects from financial text with 60% annotation level accuracyUtilized feature attribution library to interpret ML model behavior

Integrated and improved on a novel Reinforcement Learning model to learn properties of graph-based molecules Optimized ML models on an AWS cloud environment

Used Tensor Factorization and Gene Set Enrichment Analysis to observe the effects of protein-based treatments on human gene expression.

Designed and tested a Natural Language Processing Pipeline to turn event descriptions into categorical featuresCreated a Machine Learning Model to match user interest with preferred social events

• Used Machine Learning Algorithms to classify and predict trends in mortality data of several Asian subgroups and non-Hispanic whites living in the US.• Tracked Cancer and Cardio-Vascular disease deaths among Asian subgroups to motivate the disaggregation of Asian subgroups, especially in healthcare settings.• Developed a Convolutional Neural Network to detect the size of arcus senilis in corneal photographs.• Studied the efficacy of nutrition education programs in underfunded private schools in rural and urban areas in India.• Designed a curriculum about applying Machine Learning to Medical Technology

Founded and led a team of 3 to design and build an arduino based device to prevent Gravity Dependent Edema.Publication: Vedant Sathye, et al., "Testing the Feasibility of a Wearable Device to Detect Leg Position in Patients With Edema" (presentation, IEEE-EMBC 2018, Honolulu, Hawaii, August 18–20, 2018)