Home

Gregory Vance - Lead AI/ML Engineer

[email protected]

Location: Dallas, Texas, USA

Relocation: Yes

Visa: USC

Resume file: Gregory Resume_1777318471339.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Gregory Vance Email- [email protected] PROFESSIONAL SUMMARY: AI Engineer with 12+ years of extensive experience designing, developing, and deploying scalable artificial intelligence, machine learning, and deep learning solutions across diverse industry domains in the United States. Demonstrated expertise in building end-to-end AI pipelines, including data ingestion, preprocessing, feature engineering, model training, validation, and deployment in production-grade environments. Proficient in leveraging advanced AI frameworks such as TensorFlow, PyTorch, Scikit-learn, and Keras to build robust predictive and prescriptive analytics models. Extensive experience in Natural Language Processing (NLP), Computer Vision, and Generative AI applications, including large language models and conversational AI systems. Databricks & Data Intelligence: Expert in leveraging the Databricks Data Intelligence Platform (Unity Catalog, Delta Lake, and Mosaic AI) to build a unified governance layer for high-scale AI/ML, ensuring data lineage and security for sensitive healthcare and telecom datasets. AWS SageMaker Ecosystem: Proficient in AWS SageMaker for the full model lifecycle, including Ground Truth for labeling, SageMaker Pipelines for orchestration, and multi-model endpoints for cost-effective deployment of specialized domain models. LLM Integration & Orchestration: Highly skilled in integrating frontier models from OpenAI and Anthropic into enterprise applications using LangChain and LlamaIndex for advanced RAG (Retrieval-Augmented Generation) and agentic workflows. Agentic Frameworks & MCP: Pioneer in building autonomous AI Agents utilizing the Model Context Protocol (MCP) to securely bridge LLMs with private SQL warehouses, local files, and proprietary enterprise APIs. Modern Backend & Cloud Native: Expert in Python and Advanced SQL for developing scalable Microservices architectures, utilizing Docker/Kubernetes for containerization and Serverless (AWS Lambda/Fargate) for event-driven AI tasks. Production LLMOps: Specialized in building CI/CD pipelines specifically for Generative AI, focusing on automated prompt evaluation, toxicity filtering, and "LLM-as-a-Judge" scoring to ensure production reliability. API Design & Scalability: Advanced experience in designing and securing REST/gRPC APIs to serve as the backbone for high- concurrency AI systems, managing rate-limiting and token optimization across multiple AI providers. Strong background in designing cloud-native AI architectures using AWS, Azure, and Google Cloud Platform, ensuring high availability, scalability, and fault tolerance. Skilled in implementing MLOps practices, including CI/CD pipelines, model versioning, monitoring, and automated retraining for continuous improvement. Adept at working with structured and unstructured data, including big data technologies such as Hadoop, Spark, and distributed computing frameworks. Proven ability to collaborate with cross-functional teams including data engineers, software developers, product managers, and business stakeholders to deliver impactful AI solutions. Experience in developing recommendation systems, fraud detection models, predictive maintenance solutions, and customer behavior analytics. Strong expertise in RESTful API development and microservices architecture for deploying AI models into real-time applications. Proficient in Python, R, SQL, and exposure to Java and Scala for building enterprise-grade AI solutions. Hands-on experience in data visualization tools such as Tableau, Power BI, and Matplotlib for presenting insights to stakeholders. Knowledgeable in data governance, security compliance, and ethical AI practices across regulated industries. Excellent problem-solving skills with a focus on optimizing model performance, accuracy, and efficiency. Strong communication and leadership skills with experience mentoring junior engineers and leading AI initiatives at an enterprise level. AI, Machine Learning & Generative AI Core AI/ML: TensorFlow, PyTorch, Scikit-learn, XGBoost, LightGBM, JAX. Generative AI: OpenAI (GPT-4o/o1/o3), Anthropic (Claude 3.5/3.7/4.0), Hugging Face Transformers, RAG (Retrieval- Augmented Generation), AI Agents, MCP (Model Context Protocol). NLP: BERT, GPT Models, NLTK, SpaCy, LLM Orchestration (LangChain, LlamaIndex), Prompt Engineering. Data Engineering & Cloud Ecosystem Cloud Platforms: AWS (SageMaker, Lambda, S3, Fargate, EKS), Azure ML, GCP AI Platform (Vertex AI). Data Platforms: Databricks (Mosaic AI, Unity Catalog, Delta Lake, Databricks SQL). Data Processing: Spark (PySpark), Pandas, NumPy, Dask, Snowflake. Databases: SQL (Expert), MySQL, PostgreSQL, MongoDB, Vector Databases (Pinecone, Milvus, Weaviate). Engineering & Architecture Systems Design: Microservices, RESTful APIs, gRPC, Serverless Architecture, Building & Deploying AI/LLM-based systems in production. Languages: Python (Expert), Advanced SQL, R, Java, C++. Visualization: Tableau, Power BI, Matplotlib, Seaborn. DevOps & MLOps Containerization: Docker, Kubernetes (K8s). Automation: Jenkins, Git, CI/CD Pipelines (GitHub Actions, GitLab CI). Orchestration: MLflow, Apache Airflow, Kubeflow, BentoML, Ray Serve. 3M Health Care (San Antonio, TX) Oct 2020 to Present Responsibilities: Architected a production-grade RAG (Retrieval-Augmented Generation) system using Databricks Mosaic AI to enable clinicians to query 1M+ medical research papers with verifiable citations. Developed a HIPAA-compliant MCP (Model Context Protocol) Server to allow AI Agents to securely interface with patient databases and diagnostic tools while maintaining strict data governance via Unity Catalog. Integrated Anthropic Claude 3.5 & 3.7 APIs into clinical workflows to automate the summarization of complex longitudinal patient histories, reducing physician documentation time by 40%. Designed a Serverless orchestration layer using AWS Lambda and Python to trigger real-time patient risk assessments, scaling automatically to handle peak hospital admission hours. Built a microservices-based diagnostic engine housed in Docker containers and orchestrated via Kubernetes (EKS), ensuring high availability for critical AI-driven imaging tools. Established a robust CI/CD pipeline for LLM-based systems, incorporating automated "LLM-as-a-judge" evaluation steps to prevent regression in medical accuracy during model updates. Leveraged AWS SageMaker for fine-tuning specialized Healthcare LLMs, optimizing model weights for medical-specific terminology and multi-modal data analysis. Developed a HIPAA-compliant Predictive Analytics platform to identify high-risk patients for chronic disease management using historical EHR data. Implemented Natural Language Processing (NLP) models to extract clinical entities and relations from unstructured physician notes with 94% accuracy. Built a Computer Vision system for automated medical imaging analysis, assisting radiologists in detecting early-stage anomalies in X-rays. Optimized patient flow and bed management using reinforcement learning, reducing average wait times by 20% across multiple facilities. Designed a RAG-based medical assistant for clinicians to query internal medical research papers and clinical guidelines securely. Engineered an automated claims processing engine using Deep Learning to detect fraudulent billing patterns, saving the client $15M annually. Conducted rigorous bias testing on diagnostic models to ensure equitable healthcare outcomes across diverse demographic groups. Developed a real-time monitoring system for ICU patients using time-series forecasting to predict sepsis onset 6 hours in advance. Streamlined data de-identification pipelines to facilitate secure sharing of datasets for research purposes without compromising PII. Integrated FHIR (Fast Healthcare Interoperability Resources) standards into ML data pipelines for seamless data exchange. Fine-tuned BERT-based models for medical sentiment analysis to improve patient satisfaction scoring from survey responses. Managed the deployment of models onto edge devices for real-time diagnostic assistance in remote clinical settings. Led a team of data scientists to build a personalized medicine recommender system based on genomic data and patient history. Designed and implemented a cloud-native AI ecosystem on Azure, leveraging Azure Databricks, Azure Data Lake Storage Gen2, and Delta Lake to build scalable and secure data pipelines for healthcare analytics. Built end-to-end MLOps pipelines using Azure DevOps and Azure ML pipelines, enabling automated CI/CD workflows for model training, validation, and deployment with governance controls. Improved the accuracy of ICD-10 coding automation by 35% through custom-trained Transformer models. Implemented federated learning protocols to train models across multiple hospital sites while maintaining data localized privacy. Designed a proactive appointment reminder system using behavioral modeling to reduce "no-show" rates by 18%. Developed an AI-driven drug discovery pipeline to simulate molecular interactions, accelerating the early-stage research phase. Created a voice-to-text application for surgeons to document procedures hands-free with high clinical vocabulary accuracy. Established a robust MLOps framework for medical model versioning, ensuring full auditability for FDA regulatory compliance. Utilized Graph Neural Networks (GNNs) to map patient journeys and identify critical intervention points in complex care plans. Built a predictive model for hospital readmission risk, enabling personalized post-discharge care interventions. Developed a computer vision tool for monitoring surgical tool counts during operations to prevent retained-object errors. Engineered a nutrition-tracking AI that analyzes patient meal photos to provide real-time dietary feedback for diabetics. Orchestrated the migration of legacy healthcare data to a centralized cloud lakehouse for unified ML training. Mentored junior engineers on healthcare-specific data challenges, including handling missingness and imbalanced clinical datasets. Senior AI Engineer Frontier (Houston, TX) Mar 2017 to Oct 2020 Responsibilities: Built an Autonomous Network Operations Agent using OpenAI s GPT-4o API and custom orchestration logic to troubleshoot cell-site failures and execute remediation scripts via APIs. Deployed a Databricks-based Feature Store to provide real-time SQL feature serving for high-frequency churn prediction and network congestion models. Began integrating Azure Machine Learning Studio for experimentation and model prototyping alongside existing AWS-based workflows. Developed hybrid cloud solutions combining Azure Data Factory and AWS pipelines, enabling cross-platform data orchestration. Experimented with Azure Cognitive Services APIs for NLP and sentiment analysis in customer support systems. Developed a containerized Microservices architecture to handle multi-terabyte log analysis, utilizing Sagemaker Pipelines to automate model retraining based on drifting network conditions. Implemented an MCP Server integration to bridge the gap between LLM support bots and proprietary network telemetry databases, ensuring real-time technical accuracy. Engineered a Serverless "Next Best Offer" (NBO) system that utilizes Python and OpenAI to generate personalized retention scripts for call center agents in under 200ms. Orchestrated end-to-end MLOps on SageMaker, managing the deployment of distributed Deep Learning models across edge computing nodes for 5G signal optimization. Integrated multi-provider AI failover strategies (switching between OpenAI and Anthropic) within the production API gateway to ensure 99.99% uptime for customer-facing AI. Architected an end-to-end 5G Network Slicing Optimizer using Deep Reinforcement Learning to dynamically allocate bandwidth for mission-critical applications and IoT devices. Developed a Real-Time Network Congestion Prediction engine that analyzes multi-terabyte streaming telemetry data to forecast outages up to 30 minutes before they occur. Engineered a sophisticated Customer Churn Prediction model utilizing Gradient Boosted Trees (XGBoost/LightGBM) on a dataset of 10M+ subscribers, improving retention rates by 22%. Implemented an Automated Root Cause Analysis (RCA) system that processes unstructured syslog data and SNMP traps using NLP to identify hardware failures across global cell sites. Designed a Generative AI-powered Virtual Assistant for technical support that leverages RAG (Retrieval-Augmented Generation) to troubleshoot complex router and connectivity issues. Optimized Last-Mile Fiber Deployment strategies by training Computer Vision models on satellite imagery to identify geographical obstacles and optimal trenching paths. Developed a Fraud Detection Framework to identify "SIM Box" fraud and international revenue share fraud (IRSF) in real- time, saving the client $8M in annual losses. Built a Predictive Maintenance pipeline for cell tower HVAC units and battery backups, utilizing LSTMs to detect degradation patterns and prevent site downtime. Implemented a Dynamic Pricing Engine for prepaid roaming packages, utilizing elasticity modeling to maximize Average Revenue Per User (ARPU) during peak travel seasons. Architected a Geospatial AI Tool to visualize and predict signal propagation and interference in dense urban environments, optimizing 5G small-cell placement. Deployed Speech-to-Text and Sentiment Analysis models within call centers to provide real-time agent assistance and monitor compliance across thousands of daily interactions. Developed a Recommendation System for Content Bundling (OTT services/data plans) using Matrix Factorization, increasing upsell conversion by 18%. Streamlined MLOps workflows using Kubeflow and Docker to automate the retraining and deployment of signal processing models across edge computing nodes. Engineered an AI-driven Energy Management system for data centers that adjusted cooling loads based on predictive traffic volume, reducing utility costs by 12%. Developed Lead Scoring models for B2B Sales that analyzed firmographic data and usage patterns to identify high-value enterprise prospects for dedicated leased lines. Implemented Automated Image Recognition for field technicians to verify the quality of hardware installations and cable management via a mobile app. Created a Traffic Classification model using Deep Packet Inspection (DPI) and Machine Learning to prioritize latency- sensitive gaming and VOIP traffic. Designed a Federated Learning framework to improve mobile device battery optimization models without compromising subscriber data privacy. Built a Load Balancing AI for core network functions that redirected signaling traffic during flash-crowd events (e.g., major sporting events or holidays). Led the migration of legacy SAS-based analytical models to a modern, scalable Python-based environment on AWS SageMaker, reducing inference latency by 40%. Senior Machine Learning Engineer Webinopoly (Remote) April 2015 to Mar 2017 Responsibilities: Architected a multi-stage Hybrid Recommender System combining Two-Tower neural networks for retrieval and deep cross-networks (DCN) for ranking, resulting in a 28% increase in Add-to-Cart rates. Developed a Real-Time Dynamic Pricing Engine utilizing Reinforcement Learning to adjust prices based on inventory levels, competitor fluctuations, and consumer demand elasticity. Implemented an AI-driven Visual Search feature using Siamese Networks and Vision Transformers (ViT), allowing users to find products by uploading photos with a 92% Top-5 accuracy. Engineered a Lifetime Value (LTV) Prediction model using Bayesian structural time-series to segment high-value customers for personalized VIP marketing campaigns. Primarily worked on on-prem and AWS-based ML systems; explored early Azure capabilities such as Azure Storage and basic cloud-hosted compute services. Evaluated cloud platforms including Azure for future scalability but did not heavily rely on Azure ML services Designed a Generative AI solution for Automated Product Descriptions, fine-tuning LLMs to transform raw attribute data into SEO-optimized, brand-aligned copy across 1M+ SKUs. Developed a sophisticated Fraud Detection System using Graph Neural Networks (GNNs) to identify sophisticated "bot rings" and account takeover (ATO) attempts in real-time. Built a Predictive Inventory Management tool that leverages Prophet and TCN (Temporal Convolutional Networks) to forecast SKU-level demand, reducing overstock by 15%. Implemented a "Complete the Look" Recommendation engine using Deep Learning-based style compatibility modeling to increase Average Order Value (AOV) via cross-selling. Optimized Search Relevance (Learning to Rank) by implementing LambdaMART models that factor in user intent, historical clicks, and product availability. Created a Semantic Search pipeline using vector embeddings (Milvus/Pinecone) to handle long-tail queries and improve search results for complex, natural language descriptors. Engineered an automated Size & Fit Recommender to reduce return rates, utilizing historical return data and customer measurements to predict the best fit for apparel items. Developed a Customer Sentiment Engine that processes thousands of daily product reviews using Aspect-Based Sentiment Analysis (ABSA) to provide actionable insights to vendors. Architected a real-time Personalization Wrapper for the homepage that re-ranks content blocks based on the user s current session behavior within milliseconds. Deployed an AI-powered Chatbot using RAG architectures to handle complex post-purchase inquiries, such as order tracking and return policy clarifications. Optimized Logistics and Last-Mile Delivery by training a multi-agent reinforcement learning model to minimize delivery windows and fuel consumption for proprietary fleets. Implemented Automated Image Tagging and Moderation using CNNs to categorize user-generated content and filter out non-compliant imagery instantly. Developed a Propensity Model to predict "Intent to Buy," allowing the marketing team to trigger real-time discount pop- ups only for users on the verge of abandoning their carts. Built a Market Basket Analysis tool using the FP-Growth algorithm to identify hidden product associations, informing warehouse layout and promotional bundling. Established a robust MLOps Pipeline using BentoML and MLflow to manage model versioning, A/B testing, and canary deployments for high-traffic sales events like Black Friday. Designed an Attribution Model using Markov Chains to accurately distribute credit across multiple marketing touchpoints, optimizing the client s multi-million dollar ad spend. Machine Learning Engineer Gomage(Austin, TX) Mar 2013 to Apr2015 Responsibilities: Assisted in the development of a Collaborative Filtering recommendation engine, utilizing Matrix Factorization techniques to provide personalized product suggestions to a user base of 500k+. Developed and maintained automated data cleaning pipelines using Python and Pandas to process raw web-scraping data, improving the quality of training sets for price-comparison models. Implemented an image classification model using Convolutional Neural Networks (CNNs) to automatically categorize incoming product photos into 50+ pre-defined catalog departments. Conducted extensive A/B testing analysis on homepage recommendation widgets, using statistical methods to validate a 5% uplift in user engagement and click-through rates. Built a sentiment analysis tool using NLTK and Scikit-learn to classify customer product reviews as positive, negative, or neutral, providing weekly reports to the product team. Optimized SQL queries and ETL processes in Snowflake to reduce the data extraction time for model training by 30%, enabling faster experimentation cycles. Developed a baseline Demand Forecasting model using ARIMA and Linear Regression to predict weekly sales for top- performing product categories. Integrated third-party APIs for real-time currency conversion and localized pricing, ensuring the machine learning models reflected accurate global market data. Created a prototype for a "Visual Similarity" search tool, using pre-trained ResNet-50 embeddings to find similar-looking items within the footwear category. Assisted in the deployment of model microservices using Docker and FastAPI, ensuring low-latency responses for the front-end production environment. Performed exploratory data analysis (EDA) on user session logs to identify common drop-off points in the checkout funnel, informing future propensity model features. Implemented automated unit tests for machine learning scripts to ensure data consistency and model reliability during the CI/CD deployment phase. Collaborated with the DevOps team to set up basic model monitoring dashboards in Grafana, tracking drift in key performance metrics like precision and recall. Developed a rule-based fraud detection filter that flagged suspicious high-frequency transactions for manual review, reducing credit card chargebacks by 10%. Fine-tuned hyper-parameters for XGBoost models used in lead scoring, achieving a 12% improvement in identifying high- intent shoppers compared to the previous baseline. Documented technical specifications and model architectures comprehensively, ensuring smooth knowledge transfer and onboarding for new team members. Cleaned and labeled large-scale datasets for a "frequently bought together" recommendation project, utilizing association rule mining (Apriori algorithm). Built interactive data visualizations in Tableau and Matplotlib to communicate model insights and performance trends to non-technical stakeholders. Explored and prototyped NLP-based query expansion techniques to improve the accuracy of the internal site search engine for misspelled product names. Participated in weekly code reviews and sprint planning, contributing to the optimization of the overall machine learning development lifecycle. Data Engineer Alpheus Communications (Houston, TX) Jun 2010 to Sep 2012 Responsibilities: Designed and built scalable data pipelines using Python to process structured and unstructured data from multiple sources. Developed ETL/ELT workflows using tools like Airflow to automate data ingestion, transformation, and loading processes. Worked with big data technologies such as Apache Spark for large-scale data processing and analytics. Implemented real-time data streaming pipelines using Apache Kafka to handle high-velocity data ingestion. Built and optimized data models in cloud data warehouses like Snowflake and Amazon Redshift. Wrote complex SQL queries and optimized performance for large datasets, improving query efficiency and reducing execution time. Developed data validation and quality checks to ensure accuracy, consistency, and reliability of datasets. Collaborated with data analysts and business stakeholders to gather requirements and deliver actionable insights. Designed and implemented data lakes on cloud platforms such as Amazon S3. Built batch and incremental data processing pipelines to support analytics and reporting needs. Utilized dbt for transforming raw data into analytics-ready datasets. Implemented CI/CD pipelines for data workflows to ensure smooth deployment and version control. Monitored data pipelines and resolved production issues to ensure high availability and reliability. Applied data partitioning, indexing, and optimization techniques to improve performance of large-scale datasets. Worked with APIs and external data sources to ingest third-party data into internal systems. Used Git for version control and collaborated in Agile/Scrum environments for iterative development. Ensured data governance, security, and compliance by implementing role-based access control and encryption mechanisms. Built reusable data components and frameworks to standardize data engineering practices. Performed data reconciliation and troubleshooting to identify and resolve discrepancies in datasets. Automated manual data processes, reducing operational effort and improving efficiency. Documented data pipelines, schemas, and workflows to ensure maintainability and knowledge sharing. Keywords: cplusplus continuous integration continuous deployment artificial intelligence machine learning business intelligence sthree active directory rlang fiveg Delaware Texas