Experience

Leading AI/ML engineering and architecture for enterprise programs, from agentic AI systems to distributed data pipelines and GPU-optimized model training.

Achievements

• Named Agentic Architect of a newly formed enterprise agentic AI team of 6-7 engineers, leading the design and development of agentic AI systems for internal programs and departments across the organization
• Architected the team's agentic AI technology stack including Pydantic AI agent frameworks, RAG pipelines with agentic retrieval and generation, Qdrant and Chroma vector databases, custom MCP servers for database and object storage integration, and connections to self-hosted and external LLM providers
• Architected and developed custom, multi-step AI agents using LangChain and LangGraph; engineered bespoke Model Context Protocol (MCP) servers for real-time data integration with databases and object stores
• Developed a robust MLOps framework optimizing the entire ML lifecycle from model training to deployment, significantly accelerating deployment cycles and improving model reproducibility
• Leveraged PyTorch Profiler and NVIDIA Nsight Systems to diagnose inefficiencies in deep learning anomaly detection models, achieving a reduction in training times of over 94%
• Built highly distributed data pipelines with Apache Airflow and Dask using functional programming patterns, enabling scalable data processing and seamless integration of diverse data sources
• Served as principal architect of an Apache Arrow-native Data Lakehouse using Amazon S3 and PostgreSQL for efficient management of time series data, metadata, and a machine learning feature store
• Refactored a PostgreSQL database for improved normalization, performance, and scalability, reducing query response times by over 95%
• Deployed and managed a secure, scalable Kubernetes cluster on AWS GovCloud to support ML and data visualization platforms

Built Photon ML and conducted deep learning research for financial market prediction, processing decades of market data through custom high-performance pipelines.

Achievements

• Built Photon ML, an open-source machine learning framework extending TensorFlow, Keras, and PyTorch for deep learning in financial algorithms, supporting the full ML lifecycle from data preparation through model serving
• Processed 20+ years of trading and order book market data through custom Apache Arrow in-memory columnar and Parquet on-disk storage pipelines, with extensive feature engineering for technical indicators and statistical markers
• Achieved over 65% improvement in predictive accuracy for a consulting engagement using attention-based transformers and autoencoders for sporting event predictions
• Developed sequence-based predictive models for financial markets using CNNs, RNNs, Transformers, Temporal Convolutional Networks, and Bayesian Neural Networks
• Built algorithmic trading systems in Python and C++ with rules-based and ML-based models for live execution and backtesting
• Standardized codebases across TensorFlow, Keras, NumPy, Pandas, and Scikit-learn, and open-sourced multiple frameworks and libraries

Designed and deployed data pipeline architectures, ETL processes, SQL/NoSQL databases, and full-stack web applications; led technology strategy for data management across multiple clients.

Led database administration (MS SQL Server), data integrations, and web development; managed a small team of network engineers and developers.

Provided full-stack IT consulting for small to mid-sized companies; managed a team of 15 technicians and network engineers; administered SQL Server databases and client networks.