# Mario Marcolongo — AI Evaluation & Research Operations Specialist (AI / LLM Compatible Markdown) > Model Behavior · Scientific Evidence · Knowledge Integrity · Open Science & Data Quality > I evaluate model behavior, verify scientific evidence, and operate research systems. > Work Authorization: EU/EEA work-authorized; Swiss EU/EFTA permit eligible; open to employer-sponsored work authorization elsewhere; self-funded relocation considered for exceptional opportunities. > Email: me@mariomarcolongo.com | Web: https://mariomarcolongo.com | ORCID: 0000-0003-2846-7115 --- ## Executive Summary Research, evidence and AI operations specialist with eight years of auditable public-source and structured-data work, paid scientific fact-checking and editorial production, community-facing research facilitation, ownership of an open research-participation directory, and sustained adversarial testing of AI systems. On 25 July 2026, the Gray Swan Proving Ground profile displayed rank #75 (top 6%) and 110 total breaks. Entropy for Life work covers 59+ documented published projects: 55+ YouTube video projects and 4 co-authored articles. Technical implementation is described accurately as code-literate, AI-assisted operations rather than independent software development. > **I use AI-assisted implementation extensively. I define requirements and workflows, inspect code structure and behavior, test implementations, diagnose functional problems, guide iterative changes, deploy releases, and maintain services. I do not present myself as an independent software developer.** --- ## Core Competencies & Technical Skills - AI Safety Evaluation & Adversarial Testing: Exploratory model-behavior testing, prompt and jailbreak analysis, multi-turn behavior, multimodal inputs, agentic tool-use and indirect prompt injection. - Evaluation Operations: Test planning, evidence capture, reproducibility notes, taxonomy thinking, issue classification, evidence-bound reporting and mitigation-retesting concepts. - Scientific Verification & Evidence Synthesis: Primary-source fact-checking, bibliographic research, claim decomposition, evidence screening, source-quality assessment and cross-source corroboration. - Data Quality & Provenance: Structured metadata, entity reconciliation, validation rules, documentation, taxonomy design and public-record verification. - Code-Literate Technical Operations: Requirements definition, codebase reading and behavior inspection, functional testing, issue diagnosis, deployment operations and service maintenance. - AI-Assisted Technical Workflow: Uses coding agents for implementation support while personally inspecting structure and behavior, testing results and coordinating revisions; not independent software development. - Open Science & Structured Data: Wikimedia, Wikidata, FAIRsharing, Zenodo, ENA records, JSON-LD, RDF Turtle/VoID, OpenAPI and MCP interfaces. - Web & Cloud Operations: WordPress, HTML/CSS modification, Git/GitHub, JSON, REST APIs, AWS Lambda, Cloudflare Pages, DNS/SSL and CI/CD operations. - Data Visualization: SVG vector diagrams, Tableau Public, Flourish and open-data publication. - Languages: Italian — native. English — C1 overall (EF SET 68/100), with advanced technical reading and professional/technical writing. --- ## Professional Experience ### Founder & Research-Workflow Owner — Yourself to Science™ *Aug 2024 — Present* · **[Independent project]** - Founded and operate an open-source research-participation directory indexing more than 55 clinical studies, biobanks, donation programs, registries and other initiatives. - Defined the inclusion model, verification workflow, provenance fields, licensing structure and machine-readable metadata requirements. - Use AI-assisted implementation while personally defining requirements, inspecting code structure and behavior, testing releases and maintaining the service. ### Model-Behavior Evaluator — Independent practice · Gray Swan Proving Ground participant *Jul 2026 — Present* · **[Independent practice]** - Conduct self-directed testing of LLM instruction handling, policy boundaries and edge cases across chat, image, agentic tool-use and indirect prompt-injection settings. - Reached #75 on the Proving Ground (top 6%) with 110 platform-displayed total breaks on 26 July 2026; the same profile displayed Arena rank #371, 27 global unique breaks, 1,090 points and 246 submissions. - Document the visible 109/110 discrepancy and separate platform-reported outcomes from independent verification, security certification or model-wide conclusions. ### Scientific Content Quality & Operations Contractor — Entropy for Life — Italy *Jun 2023 — Present* · **[Independent contractor]** - Delivered 80 documented published content contributions: 55 YouTube videos, 4 co-authored articles and 21 short-form pieces. - Own recurring primary-literature research and scientific fact-checking; depending on the assignment, translate evidence into scripts, data analyses, visualizations, presentation slides, on-screen assets and short-form content. - Develop selected thumbnail concepts and visual packaging independently or with the video editor, using click-through rate, watch time, retention and immediate attention capture as explicit design criteria. - Designed and built entropyforlife.it in WordPress and manage responsive design, publishing, OVHcloud hosting, DNS, SSL and technical SEO; formally acknowledged in the Mondadori book Italiani veri for scientific-literature research and error detection. ### Volunteer Focus-Group Co-Facilitator & Research Operations Contributor — Department of Developmental Psychology and Socialisation (DPSS), University of Padua *Nov 2022 — 2025* · **[Volunteer research collaboration supervised by Marta Panzeri]** - Served as lead or co-facilitator across approximately 4–5 recorded Zoom focus-group sessions, typically lasting 1–2 hours, with autistic participants discussing sensitive sexuality and relationship topics. - Co-developed the protocol, including pseudonymous naming, explicit recorded consent, optional captions and written-chat participation, timed turn-taking, scripted prompts, recording boundaries and two-person facilitation handoffs. - Supported participant recruitment, bibliographic research, technical preparation and protocol feedback from an autistic perspective; coordinated with Marta Panzeri, researchers and a second autistic volunteer facilitator. - Public attribution to Marta Panzeri and the Department of Developmental Psychology and Socialisation is included with permission; no participant information or confidential session content is disclosed. ### Scientific Contributor & Structured-Data Editor — Wikipedia, Wikidata & Wikimedia Commons *Mar 2018 — Present* · **[Public contribution record]** - Completed 4,317 auditable contributions across Wikimedia projects as of July 2026. - Check citations, reconcile conflicting sources, improve structured metadata and create scientific visualizations adopted across four Wikipedia language editions. --- ## Research & Open Science Infrastructure ### Personal Genomics Workflow & Open-Data Record (41×) — European Nucleotide Archive — PRJEB109744 / SAMEA121950568 *Jan 2026* · **[Personal open-data project]** | Website: https://www.ebi.ac.uk/ena/browser/view/PRJEB109744 | GitHub: https://github.com/jnton/git-nome - Donated personal 41× whole-genome sequencing raw paired-end FASTQ reads to the public domain under ENA BioSample SAMEA121950568. - Defined analytical requirements and used Terra.bio cloud workflows to produce GRCh38 BAM alignments and VCF variant-call files. - Specified requirements for and iteratively validated a downstream workflow using Plink2 and Nextflow pgsc_calc with ancestry projection. - Used AI-assisted Python scripts for VEP-annotated VCF extraction, multi-trait polygenic-score summaries and variant filtering; inspected outputs and iterated requirements rather than independently developing the code. ### Scientific Data Visualizer & Evidence Synthesist — Wikimedia Commons, Tableau Public & Flourish *2023 — Present* · **[Independent project]** | Tableau: https://public.tableau.com/app/profile/mario.marcolongo/vizzes | Flourish: https://app.flourish.studio/@Digressivo - Published more than 70 empirical public-health, epidemiological and biomedical visualizations across Wikimedia Commons, Tableau Public and Flourish. - Created an original vector Euler diagram of overlapping monogenic clinical phenotypes adopted across four Wikipedia language editions. - Synthesized primary datasets into open-access charts, maps and vector evidence records. --- ## Deployed Systems & Open-Source Projects ### Model Behavior & Adversarial Evaluation **Model-Behavior Evaluator** Self-directed model-behavior evaluation conducted through the Gray Swan Proving Ground. The 25 July 2026 profile snapshot displays rank #75, top 6%, and 110 total breaks. The same snapshot shows 246 Arena submissions, 27 global unique breaks, 1,090 points and Arena rank #370. Aggregate counts are presented with explicit evidence limitations; complete prompts, outputs, model versions and adjudication materials are not reproduced. ### Yourself to Science™ | Open Research-Participation Directory **Founder & Research-Workflow Owner** | Website: https://yourselftoscience.org | GitHub: https://github.com/yourselftoscience/yourselftoscience.org | DOI: https://doi.org/10.5281/zenodo.15109359 | FAIRsharing: https://doi.org/10.25504/FAIRsharing.d3d487 Founded, designed and operate an open-source research-participation directory indexing more than 55 initiatives. Defined the inclusion model, verification workflow, provenance fields, licensing structure and machine-readable metadata requirements, including JSON-LD, RDF Turtle/VoID, OpenAPI and an MCP interface. Technical implementation is AI-assisted and personally verified through requirements, code reading, functional testing and operational maintenance. ### Entropy for Life — Scientific Content Quality & Operations **Scientific Content Quality & Operations Contractor** Conduct recurring primary-literature research and scientific fact-checking across 80 documented published content contributions: 55 YouTube videos, 4 co-authored articles and 21 short-form pieces. Entropy for Life is an Italian science-communication brand with 267K YouTube subscribers, 36,524,137 channel views, 159K Instagram followers and 54K TikTok followers as of 26 July 2026. Depending on the assignment, also translate evidence into scripts, data analyses, visualizations, slides, on-screen assets, short-form content and selected thumbnail concepts or production. Designed and built entropyforlife.it in WordPress and operate its responsive design, publishing and OVHcloud technical stack. The audience belongs to the brand, and cross-platform totals are not counts of unique people. ### MDPI Filter | Browser Extension **Product Owner & AI-Assisted Technical Operator** | Chrome Store: https://chromewebstore.google.com/detail/mdpi-filter/comknkeimaaadpiopddjoknflbmjeccp | Edge Store: https://microsoftedge.microsoft.com/addons/detail/mdpi-filter/efonlkldplkaeekpiajloajjmkappjgi | GitHub: https://github.com/orgs/mdpi-filter/repositories Conceived product requirements and specified DOM-targeting behavior for a browser extension that highlights or hides MDPI publications across Google, Google Scholar, PubMed and Europe PMC and identifies citations on publisher pages. Coordinated AI-assisted implementation, inspected behavior, tested releases and maintained public store deployments. ### English Wikipedia Link Converter | Telegram Bot **Creator & Technical Operator** | GitHub: https://github.com/jnton/english-wikipedia-link-converter-telegram-bot | Bot: https://t.me/ToEnWikipediaBot Specified the behavior and deployment requirements for a Telegram bot running on AWS Lambda and API Gateway. Used AI-assisted implementation, tested private, group and inline workflows, diagnosed deployment problems and maintained GitHub Actions releases. ### Emergent Humanity | Interactive Network Narrative **Creator & Concept and Requirements Designer** | Website: https://jnton.github.io/emergent-humanity/ | GitHub: https://github.com/jnton/emergent-humanity Developed the concept, narrative structure, interaction requirements and behavioral specifications for an evolving browser-based network simulation. Implementation was produced through AI-assisted workflows and personally tested and iterated. --- ## Education & Certifications - **Coursework in Medicine and Surgery** (2020) — Università degli Studi della Campania ‘Luigi Vanvitelli’ · Degree not completed - **EF SET English Certificate 68/100 (C1 overall)** (Mar 2024) — EF Standard English Test · Credential: https://cert.efset.org/en/eJz39v - **Career Essentials in Generative AI** (Mar 2024) — Microsoft and LinkedIn · Credential: https://www.linkedin.com/learning/certificates/c4f1f59578e3ac2567787e262e3b2ec55debf96bbbced4d34d5edaa821d9e6d9 - **GALENOS Crowd Evidence Synthesis Training** (May 2026) — Cochrane Crowd & GALENOS · Systematic-review screening training ## Contact & Identifiers - [Email](mailto:me@mariomarcolongo.com): Direct email contact - [Website](https://mariomarcolongo.com): Main portfolio homepage - [ORCID](https://orcid.org/0000-0003-2846-7115): ORCID scientific researcher profile - [LinkedIn](https://www.linkedin.com/in/mario-marcolongo): Professional LinkedIn profile - [GitHub](https://github.com/jnton): GitHub profile and open-source repositories - [Agent-Readiness Audit](https://isitagentready.com/mariomarcolongo.com?checks=robotsTxt%2Csitemap%2ClinkHeaders%2CdnsAid%2CmarkdownNegotiation%2CrobotsTxtAiRules%2CcontentSignals%2CwebBotAuth%2CapiCatalog%2CoauthDiscovery%2CoauthProtectedResource%2CauthMd%2CmcpServerCard%2Ca2aAgentCard%2CagentSkills%2CwebMcp%2Cx402%2Cmpp%2Cucp%2Cacp): Agent-discovery and machine-readable endpoints implemented; live audit to be rerun after deployment. - [A2A Agent Card](https://mariomarcolongo.com/.well-known/agent-card.json): Agent-to-Agent discovery card - [Genomic Pipeline Codebase](https://github.com/jnton/git-nome): Personal genomics workflow and open-data pipeline - [Portfolio SSOT Codebase](https://github.com/jnton/mariomarcolongo): Source code and Single Source of Truth for this portfolio