Selected résumé
John Cullen
Senior software engineer and independent ML researcher with experience shipping production systems and carrying long-running technical work from hypothesis through implementation, evaluation, and revision.
Independent research
Structured state, evidence binding, and conversational reasoning
- Designed a bounded active-view system that reached exact four-seed transfer on an independently authored held-out pool, with a decisive state-disabled control.
- Built evaluation workflows with preregistered gates, persisted predictions, checkpoints, negative-result records, and finite trace validation.
- Rejected natural-language front ends, bounded conversational architectures, and inadequate decoder substrates when they failed their declared promotion criteria.
- Advanced the next study to a frozen 7B decoder that passed a fresh evaluator-owned admission bank, without claiming that the attachment comparison has already occurred.
Read the bounded research case studyProfessional experience
LegalZoom
Software Engineer II → Senior Software Engineer · 2022–present
- Rebuilt checkout to support Stripe and digital wallets.
- Migrated a document-ingestion pipeline that associates state legal documents with customer entities and orders.
- Responded to production incidents and improved subscription and order-status visibility used to remediate customer issues.
Dogtown Media
Intern → Software Engineer · 2018–2022
- Built responsive React applications for external clients.
- Developed GraphQL and Express services with AWS-backed authentication.
- Built a Flask and TensorFlow API for training-data intake, media processing, training progress, and prediction workflows.
Education
Georgia Institute of Technology
Graduate study in computer science with a Computing Systems focus, including reinforcement learning, natural-language processing, algorithms, security, networks, and software analysis.
Technical focus
Python, TypeScript, Node.js, React, SQL, Docker, Linux, AWS, PyTorch, TensorFlow, Jupyter, NumPy, Pandas, LaTeX, Lean, experimental design, causal ablations, and reproducible evaluation.