--- title: "Mario Marcolongo — AI Evaluation, Information Integrity & Research" description: "Mario Marcolongo: AI red teaming and evaluation, information integrity and scientific research, in English and Italian. Paid scientific work since 2023." canonical: https://mariomarcolongo.com/ --- AI evaluation, information integrity & research # Investigate information. _Test systems._ I test AI systems and software and check the evidence behind claims, in English and Italian. Paid scientific and website work since 2023, with AI workflow setup and client training since October 2026. [View my résumé](/cv-ai) [Discuss a project](#contact) Italy · EU citizen · Native Italian, fluent English · Open to remote work & relocation Other résumés[Scientific research & fact-checking](/cv)[Software QA & AI workflows](/cv-technical) ## On the record Reviewed 31 Oct 2026 1. [13 Oct 2026 **Top 5% of Gray Swan Proving Ground** #81 · 127 platform-recorded breaks](/evidence/gray-swan-2026-10-13) 2. [Oct 2026 **Member, Syntony Research Trusted Red Team Network** Project-based network of independent specialists](/ai-evaluation#syntony) 3. [Since 2023 **Paid scientific and website work** Entropy for Life, Italian science communication](/work/entropy) 4. [Oct 2026 **Coauthor, AI Safety Formalization Atlas preprint** Five merged PRs · public on Zenodo](/work/atlas) 5. [Nov 2026 **Merged fix to Basic Memory, an open-source AI knowledge base** Agent tool output · regression test](/ai-evaluation#basic-memory) 6. [2024 **Diagram reused in a Springer Nature book chapter** Palgrave Macmillan · four Wikipedia language editions](/work/wikimedia#diagram-reuse) [Every claim, with its source and limits →](/evidence) Source quality & AI evaluation ## Contributions you can inspect. Independent adversarial model testing ### [Gray Swan testing record](/ai-evaluation) Top 5% of Proving Ground Proving Ground: #81 · top 5% · 127 platform-recorded breaks across chat, image, agent and indirect challenges. Captured 13 October 2026. Arena: 40 unique breaks · #426 · 1,430 points · 415 submissions. Captured 13 October 2026. Lucky Break: Level Up: 49th of 390 participants, with 15 platform-recorded breaks. Dated platform results. Arena and Proving Ground are separate; the evidence page preserves the displayed-total discrepancy. [Inspect the original scores →](/evidence/gray-swan-2026-10-13)[AI evaluation résumé →](/cv-ai) Proving Ground breaks by track - Chat43 of 552 - Image32 of 387 - Agent35 of 593 - Indirect16 of 547 Bar length compares the tracks with each other. [Inspect the dated scores →](/evidence/gray-swan-2026-10-13) Accepted open-source changes ### [AI Safety Formalization Atlas](/work/atlas) Five merged PRs covering source auditing, citation provenance and repository tooling, including a source-backed guide to AI agent instructions. Coauthor of the Atlas preprint, publicly available on Zenodo since October 2026. “Good work on improving sources!” [— maintainer review of PR #70 ↗](https://github.com/mbrcic/ai-safety-formalization-atlas/pull/70#pullrequestreview-5334332149) [Inspect my contributions →](/work/atlas)[Read my agent-instruction guide ↗](https://github.com/mbrcic/ai-safety-formalization-atlas/blob/66ceba19c1904f00b2a734486ee0de3f5fc5a470/docs/guide/agent-instruction-design.md) [![GitHub pull request 70 showing Mario Marcolongo's accepted source-locator contribution and merged status](/media/work/atlas-pr-70.webp)](https://github.com/mbrcic/ai-safety-formalization-atlas/pull/70) Paid scientific work · Since 2023 ## Entropy for Life My work includes primary-literature research, scientific fact-checking, scripts and articles, this map’s data collection and visualization, and designing and operating the WordPress website. I also created the data visualizations in AIRC reels that feature them; complete productions are collaborative. Client context: Entropy for Life is an Italian science-communication project covering biology, evolution and the environment, with 500K+ combined followers and subscribers across its social accounts. [Explore my responsibilities & published work →](/work/entropy)[Sharks & cannabis videos ↗](/work/entropy#video-work)[AIRC reels ↗](/work/entropy#reel-work)[Thumbnail design & direction ↗](/work/entropy#thumbnail-work) [![Interactive shark-attack map with data marks and filter controls](/media/work/entropy-shark-map.webp)My published interactive shark-attack map ↗](https://entropyforlife.it/2024/07/24/mappa-attacchi-squali-nel-mar-mediterraneo-e-in-europa/) Information integrity & investigations [Inspect the cases →](/investigations) ## Follow the evidence. I recover missing sources, reconcile conflicting records and examine scientific, privacy and corporate claims. Public revision histories show my contributions and the evidence behind them. - [Syndromic autism: building a structured biomedical evidence section](/investigations#syndromic-autism) - [Archival reconstruction of a legally sensitive public record](/investigations#archival-legal-record) - [Nebula Genomics: tracing privacy, corporate and legal claims over time](/investigations#nebula-genomics) - [Biographical source-quality and notability review](/investigations#biographical-source-review) - [Scientific & structured-data work](/work/wikimedia) Consulting & project delivery [Discuss your project →](#contact) ## How I can help. Creators, small teams and research groups. Agree the problem, deliverables, review points and handover before work starts. - ### AI assistant and agent testing Authorised adversarial testing of your chatbot or agent, in English and Italian: prompt injection, ignored instructions, wrong or unsafe answers. You get reproducible cases and a retest. [See the work: AI assistant and agent testing →](/ai-evaluation) - ### AI workflows and client training A reusable AI-assisted workflow with tool selection, operating instructions, output checks and training so your team can run it independently. [See the work: AI workflows and client training →](/work/entropy) - ### Scientific fact-checking and source investigations Trace claims to original sources, assess evidence quality and support citation audits, research and publishing. [See the work: Scientific fact-checking and source investigations →](/investigations) - ### Data visualisation and research support Organise and check source data, create interactive charts and explain findings with their provenance and limitations. [See the work: Data visualisation and research support →](/work/scientific-visualizations) - ### Software and localisation QA Manual and regression testing, English-Italian interface checks and reproducible bug reports for websites, apps and browser tools. [See the work: Software and localisation QA →](/notandia) - ### Agent instructions and prompt engineering Maintainable instructions, task routing and checks for AI coding agents, based on platform guidance and context limits. [See the work: Agent instructions and prompt engineering →](/work/atlas) - ### Website delivery and workflow improvement Deliver or improve websites and publishing workflows, diagnose issues, verify changes and document the setup for ongoing use. [See the work: Website delivery and workflow improvement →](/work/entropy) Independent products [All projects →](/work) ## Tools for research. [![Actual Notandia settings with configurable MDPI and Frontiers watchlists and optional research-integrity checks](/media/work/notandia-current-options.webp)](/notandia) Product requirements · Testing · Release work ### [Notandia](/notandia) Configurable publisher watchlists and explainable research-integrity signals. Screenshot from the current source build. [My contribution →](/notandia)[Source & installation ↗](https://github.com/notandia/browser-extension) [![Yourself to Science directory showing search filters and research-participation resources](/media/work/directory-live.webp)](/work/yourself-to-science) Founder · Source curation · Maintenance ### [Yourself to Science](/work/yourself-to-science) A directory of research-participation opportunities with filters, inclusion decisions and original-source links. [My contribution →](/work/yourself-to-science)[Try the directory ↗](https://yourselftoscience.org/) Published visual explanations [Full collection →](/work/scientific-visualizations) ## Make the evidence visible. [![Benzo(a)pyrene contamination measured across commercial brands and sampling years, synthesized from published HPLC and GC-MS toxicology literature.](/media/work/wikimedia-yerba.svg)Carcinogen Levels in Yerba Maté](https://commons.wikimedia.org/wiki/File:Benzo\(a\)pyrene_Concentration_in_Processed_Yerba_Mat%C3%A9_Leaves_Sampled_in_2006,_2008,_and_2010_-_Column_Chart.svg) [![Comparative map of food-based dietary guidelines and their positions on vegetarian diets. Sources and dates are documented with the original artifact.](/media/work/wikimedia-vegetarian.svg)Global Vegetarian Diet Policies](https://commons.wikimedia.org/wiki/File:Countries_\(States_and_Subnational_Regions\)_and_Their_Positions_on_Vegetarian_Diets_in_Food-Based_Dietary_Guidelines.svg) Scroll sideways to preview more charts. Original Wikimedia Commons diagrams. Tableau maps, Flourish stories and further data views remain in the full archive. ## Assessed skills Microsoft · Earned 19 October 2026 ### [Microsoft Applied Skills: Build an agent in Microsoft Copilot Studio](https://learn.microsoft.com/en-us/users/mariomarcolongo/credentials/48386317f9899eb1) Passed an interactive hands-on lab assessment covering agent configuration, knowledge and generative answers, conversation topics, tools and sharing in Copilot Studio. Credential ID: 48386317F9899EB1 · 4 related Microsoft Learn training modules completed. Atlassian · Earned 24 October 2026 ### [Atlassian Credentialed in JSM Fundamentals with AI](https://cp.certmetrics.com/atlassian/en/public/verify/credential/4ec41335733f49508b42ab3faab010d1) Passed the Jira Service Management with AI Fundamentals exam (ACH-815), covering app navigation, work in service spaces and request resolution. Credential ID: 4ec41335733f49508b42ab3faab010d1 · Valid through 24 October 2027 Experience & education [Full experience & responsibilities →](/experience) ## The wider background. Since 2018 · Public contributions ### [Wikipedia, Wikidata & Wikimedia Commons](/work/wikimedia) English-to-Italian translation, citation review, multilingual metadata and scientific visualizations. Nov 2022–approximately Jun 2025 · Remote volunteer collaboration, Italy ### [University of Padua research project](/research-operations) Focus-group facilitation and accessible research procedures under Marta Panzeri, Department of Developmental Psychology and Socialisation (DPSS), with a fellow facilitator. Aug 2026–Present · Open UAS ### Metropolia University of Applied Sciences Open UAS IT path studies toward a BEng. Python Programming: 3 ECTS, Pass (September 2026). Continuing Python, SQL and quantitative foundations. Further explorations [![Protein-supply explorer displaying its country map and time-series interface](/media/work/protein-explorer.webp)](https://mariomarcolongo.github.io/protein-by-bodyweight-country/) ## Protein / kg explorer Explore national protein supply relative to estimated bodyweight. Sources and model assumptions stay visible. [Explore the data ↗](https://mariomarcolongo.github.io/protein-by-bodyweight-country/)[Hypermandala source & scope →](/work/hypermandala)[Telegram & cloud automation →](/work/telegram)[More projects & contributions →](/work) ![Portrait of Mario Marcolongo](/media/mario-marcolongo-portrait.webp) International opportunities ## Discuss a role or project. Available for employed roles, contracts and scoped projects in AI evaluation, information integrity, research, AI workflows and software QA. Share the problem, intended users, desired outcome, timeline and indicative budget. [me@mariomarcolongo.com](mailto:me@mariomarcolongo.com) AI evaluation & red teaming Information integrity & research AI workflows & software QA Please email first to arrange a call. Consultations and calls are by appointment. Afternoon and evening appointments are usually available (Europe/Rome); other times by arrangement. Available for international remote work and on-site projects by prior arrangement, with travel scope and expenses agreed before booking. Open to relocation and international remote work. I can work in the EU as an Italian citizen; opportunities elsewhere depend on the role’s hiring and work-authorization arrangements. [View application résumés →](/cv)[Book through Malt ↗](https://www.malt.com/profile/mariomarcolongo)