• Led Sagely's ML pipeline development, enabling semantic search, recommender systems, and chatbots. Implemented pipelines on AWS utilizing Metaflow. Utilized Zyte for robust metadata retrieval and LangChain with OpenAI's LLMs for summarization, keyphrase extraction, full-text embeddings, and RAG for an enhanced AI chatbot. This comprehensive technological approach was designed to provide seamless coaching support aimed at boosting customer engagement.• Established Sagely's ML team, integrating advanced tools for initial workflows and capabilities. Led and fostered collaborative development of innovative ML projects across business domains.

◦ Directed a team of 5, developing and implementing a rule-based classifier, which automated the security classification and marking of various document types, contributing to enhanced data security protocols.◦ Conducted comprehensive research on embedding quality, leveraging hyperparameter sweep and exploring different techniques across multiple LMs, establishing optimal clustering pipeline, and enabling integration of NLP-based classifiers with rule-based system.

◦ Orchestrated the design and deployment of a secure, AI-optimized $3.5M private cloud, utilizing OpenStack, Kubernetes, and Nvidia A100s, enhancing GPU compute across MITRE.◦ Assembled and managed specialized AI team of 5-7 engineers, delivering “as-a-service” compute, storage, and security capabilities, expediting MITRE's AI adoption, and mitigating counter-AI vulnerabilities across $900M in customer investments.◦ Executed a rigorous evaluation and integration of workflow orchestration tools, enhancing MLOps workflows and augmenting the AI Platform's efficiency and technical capabilities.◦ Guided AI platform adoption among high-profile organizations within MITRE, optimizing client outcomes through effective technical solutions and maximizing the utilization of the platform's state-of-the-art capabilities.

◦ Led a dedicated team of approximately 10 professionals, steering the development and delivery of tools to assess adversarial ML attacks on US Government object detection models. Innovated a novel watermarking approach for dataset security and model training verification, supporting robust and secure ML practices.◦ Engineered and maintained a comprehensive suite of tools that automated the training, evaluation, and inference across hundreds of SOTA object detection models, empowering MITRE teams, including those with limited computer vision expertise, to effectively train and inference models with both online and offline capabilities.◦ Integrated hyperparameter tuning into tooling, enabling the strategic setting of intelligent defaults and facilitating the fine-tuning of adversarial attacks on diverse models and data, enhancing the adaptability and performance of attack simulations. ◦ Successfully transitioned red team's tooling to US Government infrastructure, ensuring secure analysis, reporting, and performance evaluation on sensitive data and models, thereby aligning with stringent government data handling and analysis protocols.

◦ Architected and deployed ML tools for automated training, evaluation, and inference of transformer-based Named Entity Recognition models, enhancing deidentification of sensitive safety report data and enabling model creation by teams without NLP expertise.◦ Invented a deep learning anomaly detection tool, employing hyperparameter tuning, SOTA transformer model embeddings, and an autoencoder, which identified 92% of anomalous safety issues and revealed 2 systemic issues within top 25 results, contrasting with 20% and 0 respectively, via random sampling.◦ Crafted NLP tools automating fine-tuning of numerous SOTA Transformer-Based Language Models and enabling the training, evaluation, and inferencing of text classifiers, achieving an average of 10.24 F1 improvement across multiple highly unbalanced aviation datasets by employing adaptive sampling strategies and loss function weighting.◦ Collaboratively developed and deployed both traditional and deep learning-based predictive ML models, accurately forecasting unstable approaches 24 hours in advance through meticulous data processing and cleansing via PySpark.

◦ Engineered and implemented a robust, scalable image retrieval and automated tagging method, utilizing CNN feature vectors and employing dimensionality reduction, alongside an Approximate Nearest Neighbor search, ensuring swift, efficient image data retrieval and utilization.◦ Strategically established a mirrored US Government system infrastructure using OpenStack Object Store and MongoDB, facilitating accelerated internal development by ensuring compatibility and seamless integration with government systems upon deployment.

◦ Spearheaded the standardization of Metroplex Post Implementation Analysis employing PySpark/TDP, drastically curtailing data processing times from days to hours, while elevating data accuracy and ensuring reproducibility across workflows.◦ Executed comprehensive data analyses utilizing PySpark, skillfully managing diverse data types and storage mechanisms, and adeptly presenting insights and results through Jupyter and PowerPoint, catering to various stakeholder needs.

◦ Collaboratively developed the company's strategic roadmap and business plan, providing technical direction and overseeing contract negotiations with industry leaders like General Atomic.◦ Established a secure network infrastructure, incorporating end-to-end encrypted communication and collaboration tools to facilitate secure and efficient business operations.

Multistage Localization for High Precision Mobile Manipulation Taskso Developed a novel localization approach capable of achieving sub-millimeter accuracy without having to recalibrate the position of the robot's end-effector each time the robot's mobile base moveso Implemented multistage localization approach for high precision drilling and sealant application operations through the development of autonomous navigation, manipulation, computer vision, and task management systemsSelf-Driving Vehicle Team (SDVT)o Co-founded and led a student design team focused on the development of a fully autonomous vehicle for research and development purposes, fostering innovation and collaboration among a team of 10+ students.

Relevant Data Science CourseworkIntermediate Data Analytics & ML and Intro to Data Science (R and Python)o Compared different regression models (linear, dim. reduction, bagged/boosted trees, random forest, and Bayesian) performance in predicting subway ridershipo Acquired and cleaned both weather and subway ridership data, ~10 MB datasetComputer Vision (Python)o Utilized K-means clustering to match video frames given bag-of-words representation of the selected frame or user-selected region, ~5.5 GB dataseto Used K-NN to analyze motion history images, ~65 MB, to classify human actionsAdvanced Applied Bayesian Statistics (R and VBA)o Used Gibbs Sampling to predict the number of hurricanes per montho Cleaned and converted hourly NOAA weather data, ~15 MB, to monthly weather data

Fluid Systems and Ventilation Engineering (10-week internship)o Worked on design and maintenance of the chill and potable water systems, as well as the HVAC systems on aircraft carriers

Automated Leave Tracking Systemo Designed and created a Regimental leave tracking system for 1000+ cadets using Google Forms and Sheetso Created a fillable PDF leave request forms to replace 60+-year-old paper copieso Began the development of an Access database to track and approve the created PFD requestsRegimental Survey Feedbacko Cleaned Regimental survey data from 1000+ cadets and created metrics in order to give customized feedback to Regiment, Battalion, and Company level staff about their unit’s healthSquad Tactical Challenge Scoring Systemo Designed a tracking and scoring system using Google Forms and Sheets to score 110 squads(teams) in real time; saving approximately 175 man hours Regimental Bookshareo Created a searchable Access database to facilitate the sale of textbooks between cadets

ME 4015, Engineering Design and Project | SWAT Boto Led team of 10 to design and build a surveillance robot for Virginia Tech’s ERTME 2024, Engineering Design and Economics | AeroChargeo Led team of 3 to design a hitch assembly for recharging a stationary electric vehicle

Responsible for the safety, well-being, and development of Echo Company, approximately 85 undergraduate cadets.

Test Engineering (10-week internship)o Created an Access database to assign, email, and track procedure updateso Exposed to the holistic nature of engineering: budgeting, scoping, training and the testing of systems on nuclear submarines and aircraft carriers

Guided 6 General Engineering Freshmen through their first semester at Virginia Tech

Responsible for making sure that the customer was taken care of in a timely and professional fashion, cleaning the shop, and counting the registers till fund.

Supervised 10 employees and managed a specialty bakery in the park.Oversaw revenue and labor utilization, as well as, placed orders; performed scheduling, monitored food quality, and ensured customer satisfaction.

Responsible for making sure that the customer was taken care of in a timely and professional fashion, cleaning the shop, and counting the registers till fund.

Responsible for making sure that the customer was taken care of in a timely and professional fashion, cleaning the shop, and counting the registers till fund.