PhD Astrophysics, University of Florida (2021). 5 years industry as Staff MLE until January 2026, before focusing on shipping AI safety infrastructure full-time on personal savings.

Active projects

- mcp-test-toolkit: open-core MCP server testing framework. MIT-licensed core (pytest plugin, statistical primitives with published-reference citations, protocol conformance suite) ships before July, in time for PyCon Colombia 2026 in July. Post-1.0 direction is a runtime guardrail proxy plus observability dashboard on the proprietary commercial side.

- Verity: MCP server implementing the First Principles Framework for AI-assisted decision-making. The framework formalises the Weakest Link bound and Gödel t-norm conservative aggregation principles published in my ICLR 2026 workshop papers.

- A runtime sandbox for agentic-AI execution. Currently in private development.

Recent papers and talks

- Structured Abductive-Deductive-Inductive Reasoning for LLMs via Algebraic Invariants: ICLR 2026 Workshop on LLM Reasoning. OpenReview: https://openreview.net/forum?id=zy6HdcsJ9V

- Epistemic Accountability for Agentic Financial AI: The Transformer Mandate and Evidence Lifecycle Management: 2nd ICLR Workshop on Advances in Financial AI (AFA), Paper #96.

- ACL-GEM Workshop paper, accepted April 28, 2026.

- PyCon Colombia 2026 talk: Your AI Eval Is Lying To You: Statistical Rigor for Non-Deterministic Models. Submitted April 28.

Applications in flight

Constellation Astra Fellowship (submitted), OpenAI Safety Fellowship (deadline May 3), MATS Autumn 2026 (deadline early May), LTFF, OpenPhil CDTF.

Approach

I'm building eval infrastructure outside the major labs. The work is intended to be deployable independently of any particular lab's incentives.

GitHub: github.com/astrogilda