Full-Stack & AI Engineer. I build production-grade intelligent systems — from LLM pipelines and RAG architectures to enterprise-scale platforms that ship to real users.
I'm Don Marckheil Desmond — a Computer Science, Data Science & Statistics student at UMass Amherst, graduating in 2028. I build things at the intersection of full-stack engineering and AI, with a focus on systems that go live and handle real users at scale.
From training ML classification models that serve 50,000+ employees at AbbVie, to shipping a RAG document intelligence platform at a hackathon that got adopted into production — I care about engineering rigor, measurable impact, and the craft of building AI systems that are fast, accurate, and reliable.
Outside of work I stay sharp on quantitative finance and algorithmic trading, and I'm always exploring what's next at the frontier of AI tooling and infrastructure.
Led a team of 4 building an end-to-end RAG-based document analysis system for pharma use cases — clinical trial documents, regulatory filings, and R&D reports exceeding 1,000+ pages. Architected a hybrid search pipeline combining BM25 sparse retrieval and semantic search over ChromaDB with Cohere reranking. Integrated Claude Vision for figure understanding at sub-320ms page load latency.
Predicts post-earnings stock direction (T+1/T+5/T+20) by extracting LLM signals from 500+ SEC 10-K/10-Qs across 60+ tickers. Engineered an LLM extraction layer using Anthropic tool use to pull 8 structured signals per filing for under $0.80 total. Trained a GradientBoostingClassifier with TimeSeriesSplit cross-validation to prevent look-ahead bias.
Full-stack AI clinical documentation system — Whisper speech recognition + GPT-4o mini converts clinician voice and text into structured SOAP notes, cutting documentation time by 40%. Schema-guided generation with JSON validation and retries reduced hallucination rates and latency by 40%. Deployed on Vercel with CI/CD, cutting OpenAI costs by 25%.
I'm always open to talking about internships, research, collaborations, or just interesting problems at the frontier of AI and software engineering.