AI systems builder focused on model training, agentic infrastructure, local-first intelligence, rigorous evaluation, multimodal systems, and governed automation.

Hugging Face portfolio · Microsoft Learn profile · Verified Microsoft AI-900 · ORCID · AI field notes · Email

• AgentReady v1.0.1 — public GitHub release — an evidence-driven system for governed agentic development: task coordination, tool routing, quality gates, release discipline, and verifiable handoffs.

I publish performance claims only when the supporting evidence and artifacts are ready for inspection.

• Agentic and local-first systems — model routing, tool use, memory, observability, and failure handling under real resource and authority constraints.
• Model curriculum, data, and training workflows — structured learning design, dataset preparation, adaptation experiments, provenance, and evaluation.
• Evaluation and reliability — held-out capability exams, regression analysis, evidence-backed promotion gates, and honest failure reports.
• Governed automation — deterministic controls around model reasoning so high-impact actions remain observable and human-authorized.

The scientific method is the operating model: define the hypothesis, instrument the system, challenge the result, report the limits, and promote only what the evidence supports.

• SagitariusAI — AI Systems Portfolio
• Microsoft Certified: Azure AI Fundamentals (AI-900) — online verifiable
• Microsoft Learn — 85 badges, 13 learning paths, and 117,975 XP
• The Loss Curve Said Mastery. The Exam Said Nothing Was Learned.
• Can Small AI Models Reliably Choose Tools?
• Does a Token Earn Its Keep?
• The Case of the Missing Tokens

Open to AI engineering roles, technical collaborations, research partnerships, and carefully scoped client work.

Contact: alex.burca@gmail.com