Hi, I'm Matthew McDowall
I build end-to-end AI systems that ship to production — not just demos. From agentic LLM workflows and RAG pipelines to the data infrastructure under them, I turn ideas into products people actually use. Got an idea?
About
I'm Matthew McDowall — an applied AI engineer who builds custom, domain-specific, end-to-end AI applications. Not demos. Full systems that go from client conversations and concepts to production.
I'm the Applied AI & Automation Engineer Lead at North Light AI, where I went from intern to owning the design and production rollout of our core products: an agentic supplier-matching platform used by 200+ SMBs and NASA, and Lesson Desk, an AI teaching platform. B.S. in Data Analytics & Data Science from the University of New Hampshire, class of 2026.
Exploring models, datasets & spaces
Tech I Work With
Things I've Built
Prime Ready — NHADC
Agentic LLM matchmaking platform that pairs small-to-medium suppliers with Prime contractors through natural-language search, cutting 6+ weeks of manual discovery to minutes. Concept to production in 2 months with a 4-person team; now reaching 200+ SMBs and a NASA deployment. Tuning the hybrid retrieval stack raised hit rate@5 from 74% to 92%.
AI-Driven NMS Chatbot — SubCom
Locally hosted assistant for SubCom's subsea cable Network Management System. Parses proprietary NMS documentation with Docling, embeds it into ChromaDB, and answers natural-language questions grounded in the source docs — 94% retrieval accuracy (P@10, NDCG@5), sub-second latency on VRAM-constrained VMs, fully air-gapped with zero external API calls.
Lesson Desk — North Light AI
AI-powered teaching platform that automates lesson planning and content creation so educators get their time back. Leading development from architecture through production: Claude-driven generation, a rich in-browser editor, and export to the formats teachers actually use.
Experience
Applied AI & Automation Engineer Lead
Intern (Dec 2024) → part-time engineer (Sep 2025) → full-time engineering lead (May 2026), owning the design and production rollout of North Light AI's core products and managing interns.
- Architected and deployed a fully agentic LLM system on AWS with Python and LangGraph, cutting supplier discovery from 6+ weeks of manual work to minutes via natural-language search
- Drove concept-to-production delivery in 2 months with a 4-person team, reaching 200+ SMBs and a NASA deployment
- Leading development of Lesson Desk, an AI-powered teaching platform that automates lesson planning and content creation for educators
- Raised retrieval hit rate@5 from 74% to 92% on the supplier-profile RAG pipelines through pgvector embedding-model evaluation, chunk-size tuning, and hybrid vector-search scoring
- Cleaned and integrated ~100 GB of messy, unstructured data sources via REST APIs, with Git-managed codebases supporting reliable iteration and releases
Student Intern
- Built an LLM-powered co-advisor tool end to end under a Google-funded grant: a RAG pipeline over the advisor knowledge base returning grounded, context-aware answers, moving advisors from a 2:1 to a 1:1 advisor-to-client ratio
- Worked directly with SBDC advisors across the country to identify client-support best practices that shaped the tool's design and feature prioritization
Data Analyst Intern
- Engineered an automated data-processing pipeline for weekly coastal measurements using Python and scheduled cron jobs, cutting manual staff effort by ~10 hours/week
- Designed data-validation protocols with ArcGIS and Python that improved measurement accuracy by 25% and streamlined data access for researchers and external stakeholders
Member, Reinforcement Learning Research Lab
- Weekly seminar presenting and peer-reviewing current reinforcement learning research literature
Education
B.S. in Data Analytics & Data Science
Relevant Coursework: Algorithms, Machine Learning, Linear Algebra, NLP, Reinforcement Learning, Information Retrieval, Probability & Statistics, Game Theory