AI model-behavior evaluation
Adversarial and edge-case testing across chat, multimodal inputs, tool use and indirect prompt injection.
AI evaluation · scientific fact-checking
I test AI systems and verify scientific claims.
My primary work is AI model-behavior evaluation and scientific fact-checking. I also build and operate research-information tools when a recurring verification problem needs a practical system.
Italy · EU work authorization · worldwide relocation · international contracting
Professional scope
Where I can contribute.
Scientific fact-checking
Primary-literature review for scripts, articles and public scientific communication.
Research information systems
Structured records, provenance rules and maintained tools for recurring research-information problems.
Selected work
Selected work, shown through the actual output.
Each case shows the actual public output, my contribution and the boundary of what that work demonstrates.
01AI evaluation
Finding model failures across four evaluation surfaces.
Across four public evaluation surfaces, I test chat, image, agent and indirect prompt-injection behavior and preserve dated evidence of the platform-reported result.
- What I did
- I vary the interaction path, preserve reproduction notes and distinguish direct observations from platform labels, independent verification and model-wide conclusions.
- Result
- #75 on the Proving Ground, top 6%, with 110 platform-displayed total breaks on 26 July 2026; the Arena profile displayed rank #371, 246 submissions, 27 global unique breaks and 1,090 points.
Scope: The four visible area counters sum to 109 while the profile displays 110 total breaks. Both are reported without inferring the platform’s internal aggregation. This supports evaluation and adversarial-QA applications, not penetration-testing or senior red-team engineering claims.
Research, content and website work
I contributed to 80 documented published pieces.
80published contributions55 YouTube videos · 4 articles · 21 short-form pieces
Research and fact-checkingPrimary literature, source verification and evidence synthesis
Content productionScripts, data analysis, visualizations, slides and selected thumbnails
Website design and managementWordPress, responsive design, publishing and OVHcloud operations
YouTube channel
36.5Mviews
Official evidence indexVideos · articles · short-form work · selected thumbnails
View the official Entropy for Life work record↗02Scientific content quality & operations
Evidence quality and content operations at creator scale.
Paid contractor supporting an established Italian science-communication brand across evidence review, content production and website operations.
- What I owned
- Recurring primary-literature research and scientific fact-checking. Depending on the assignment, I also developed scripts, data analyses, visualizations, slides, on-screen assets, short-form content and selected thumbnails. I designed and built entropyforlife.it in WordPress and manage its responsive design, publishing and OVHcloud technical operations.
- Result
- 80 documented published content contributions: 55 YouTube videos, 4 co-authored articles and 21 short-form pieces. The official work record also indexes selected thumbnail work, which overlaps with video projects and is not added to the total.
Scope: Platform metrics describe the production environment, not a personal audience. Quantified thumbnail lift is stated only when comparable analytics are available.
03Supporting research system
Building a maintained directory of research opportunities.
Yourself to Science turns scattered institutional opportunities into a public catalogue with explicit inclusion and update rules.
- What I did
- I defined the inclusion criteria, verification fields, provenance model, licensing boundaries, update workflow and public-data requirements, then coordinated AI-assisted implementation and ongoing operation.
- Result
- More than 55 opportunities indexed, with FAIRsharing and Zenodo records and human- and machine-readable interfaces.
Scope: My contribution covers requirements, information architecture, verification, functional testing, deployment diagnosis and operations—not unaided software development.
Current product
MDPI Filter now works in the browser and as a Zotero plugin.
The current product identifies MDPI references across literature-search and reference-management workflows while avoiding ambiguous title-based matches. The broader rebrand and expansion to retractions, comments and other research-integrity signals are future work, not shipped functionality.
Chrome · Edge · Firefox · Safari source · Zotero 7–9
Scientific visualization
A diagram that became a reusable public reference.
I designed this vector diagram to show overlap among monogenic conditions associated with autism, dystonia, epilepsy and schizophrenia. The Wikimedia record exposes the source file, authorship, revision history and reuse across four Wikipedia language editions.
Open the Wikimedia source record,_dystonia,_epilepsy_and_schizophrenia.svg)
More public work
Additional tools and experiments.
Working preferences
Clear goals, direct feedback and responsibility for the result.
I work best when the objective and decision rights are explicit, feedback is direct and I can follow a problem through investigation, documentation, release and maintenance.
Separate observation from interpretation
State what was directly observed before drawing a broader conclusion.
Make assumptions visible
Document the definitions, exclusions and judgments that affect the result.
Plan for maintenance
Treat updates, provenance and operational recovery as part of the work.
Experience
Relevant experience.
The targeted CVs carry the full application detail.
Model-Behavior Evaluator
Independent practice · Gray Swan Proving Ground participant · Conduct self-directed adversarial testing across chat, multimodal, agentic tool-use and indirect prompt-injection settings.
Scientific Fact-Checking, Content Operations & Audience Optimization Contractor
Entropy for Life — Italy · Delivered 80 documented published content contributions—55 YouTube videos, 4 co-authored articles and 21 short-form pieces—for an Italian science-communication brand with 267K YouTube subscribers and 480K+ combined platform following; recurring evidence review, content production, selected thumbnails and website operations.
Founder & Research-Workflow Owner
Yourself to Science™ · Founded and operate an open-source research-participation directory indexing more than 55 initiatives, with documented verification, provenance and metadata workflows.
Application documents
Start with the role you are hiring for.
The AI evaluation CV is the recommended default. The others are deliberately tailored alternatives.
Recommended for AI evaluation and safeguards roles
AI Evaluation & Model Behavior
Model-behavior testing, adversarial QA, evaluation operations and evidence-bound reporting.
Open CV →Scientific AI and research-data rolesScientific AI Quality & Research Data
Scientific evidence review, research-data quality, provenance, metadata and research operations.
Open CV →Integrity and investigation rolesTrust, Safety & Knowledge Integrity
Source provenance, public-record research, content integrity and investigation support.
Open CV →Editorial, content-operations and audience-quality rolesResearch, Editorial & Content Operations
Scientific fact-checking, editorial coordination, audience packaging, evidence synthesis and content operations.
Open CV →Current focus
AI evaluation and scientific evidence roles.
Especially model-behavior evaluation, scientific fact-checking, research-data quality and knowledge-integrity work.