Ann Arbor, Michigan

• • Contributing to an open source mobile AI inference engine optimized for smartphones and low power devices
• • Implementing ARM specific SIMD kernels for matrix multiplication and attention mechanisms with KV cache quantization
• • Developing zero copy computation graphs for efficient on device transformer inference with sub 50ms time to first token
• • Building OpenAI compatible FFI APIs for cross platform SDK integration across Flutter, React Native, and Kotlin

Ann Arbor, Michigan

• • Engineering a dual inference AI assistant for NASA SUITS and RASC AL challenges, integrating lightweight transformers (TinyBERT) on Jetson Orin Nano via ONNX Runtime for sub 350 ms inference, including troubleshooting deployments
• • Designing an edge cloud architecture that routes inference based on query complexity, latency, and network state
• • Implementing DSPy based orchestration and Pydantic validated NLP pipelines within Unity Inference Engine
• • Integrating WebSockets and Whisper based preprocessing to synchronize telemetry in high noise environments

Dublin, Ireland

• • Built a startup in 7 weeks as part of the Patch Summer Accelerator, leading end to end technical full stack development
• • Offsited at Stripe Dublin and developed the mobile frontend using React Native, optimizing UI responsiveness and UX
• • Implemented a scalable backend with Django and PostgreSQL to manage health data, authentication, and notifications

Remote

• • Maintained CI/CD pipelines on AWS, automating deployments across 12+ microservices and improving reliability by 28%
• • Developed Python scripts to streamline backend processes, reducing manual overhead by 40% and saving 15+ hours weekly
• • Collaborated with senior devs to optimize cloud infrastructure, achieving 35% cost reduction through resource management
• • Deployed cloud native solutions with AWS CI/CD tools, improving system scalability and achieving 99.7% uptime

Ann Arbor, Michigan

Cork, Ireland

• • Developed a backend microservice to host ML models through Flask REST APIs, for model loading, inference, health checks
• • Handled inference requests using RabbitMQ for scalable task queuing, enabling reliable processing of high-volume workloads
• • Persisted model outputs and user queries in MongoDB for analytics and logging, and containerized the system using Docker

• • Built an end-to-end Neural Radiance Fields (NeRF) pipeline for photorealistic 3D reconstruction from multi-view images
• • Implemented volumetric ray marching, positional encoding, and differentiable density/color prediction for view synthesis
• • Designed a modular training pipeline with configurable hyperparameters to support multiple 3D reconstruction datasets