# Arquestra > Arquestra is Pablo Ramos-Sánchez's archive: the tools he developed, the posts, data and white papers he publishes on how Europe adopts and governs AI, and Arquestra Chem, his R&D product (2025 to 2026). He started it in May 2025; everything on arquestra.dev is free and open. Facts in this file come from https://arquestra.dev/data/arquestra.json, the same knowledge file the site's own assistant uses. Updated 2026-10-30. ## Who - Pablo Ramos-Sánchez works on AI adoption and evidence. He has spent years taking AI into industrial R&D, and more and more of his work is about how Europe adopts and governs AI. PhD in Chemistry (KU Leuven × Covestro, industry-funded); built AutoReact, an R&D platform adopted inside Covestro (2019 to 2022); first hire at Exomatter, with industrial clients such as Airbus (2023); Arquestra Chem, AI agents for computational chemistry, brought to a large chemical company's R&D team (2025 to 2026); taught photochemistry at Goethe University Frankfurt. Based between Madrid and Frankfurt am Main. - Now: open to roles in AI policy, governance and adoption, and to consulting projects. Contact: LinkedIn, https://www.linkedin.com/in/pablo-ramos-s%C3%A1nchez - Arquestra on LinkedIn: https://www.linkedin.com/company/arquestra-compchem/ ## Tools he developed (free, in the browser) - [Consulting Market Radar](https://arquestra.dev/assets/consulting): public consulting and digital-consulting projects across Europe from EUR 15,000 up (TED with eForms details, and national registers), 24 months of awards and the calls open now, in English, Spanish, French, German and Italian: who wins, how many bid, price weight and discounts on the estimate, what comes back in the next six months. Updated every week. - [AI Procurement Observatory](https://arquestra.dev/assets/ai-observatory): search 3.7 million public purchases (461,877 in IT, plus health and more) from six official European procurement registers (TED, PLACSP, ANAC, DECP, Datenservice Öffentlicher Einkauf, TenderNed) by meaning, in any language, with every purchase that mentions AI labelled AI or not by the method of the white paper Unseen AI (reviewed or estimated, with the precision measured). Free, for research. - [AI pilot case](https://arquestra.dev/assets/ai-pilot): a 4-minute playable case on why AI pilots stall between demo and production. - [The pink elephant in the box](https://arquestra.dev/assets/audit): AI governance in practice. A public audit of data curated with AI before any number is published from it: about a hundred checks in eleven families (origin, coverage of the space, curation of the rows, what the AI decided, statistics, time series, rankings, groupings, relations, the published sentence), grouped in nine questions (universe, selection, variable, comparable, one value, rule, duplicates, time, circular); a check that applies but has no data shows hollow, a hand-checked random sample with its 95% range, and a traffic light per number. First case: the AI Project Radar. Engine: https://arquestra.dev/audit/engine.py (Python, standard library). MCP server for agents: https://arquestra.dev/api/mcp/audit (Streamable HTTP, no key; tools start_audit, measure_in_app, prepare_script, show_audit, example_audits). The person's data never reaches the server and nothing is stored. ## Reading - [Research](https://arquestra.dev/research): white papers, open access. First: Unseen AI: what Europe's public sector buys, and what its procurement registers cannot show (v1.2, 2026, CC BY 4.0, https://arquestra.dev/papers/unseen-ai.pdf), a study of 14.5 million European public purchase records; submitted to Data & Policy (Cambridge University Press). - [Insights](https://arquestra.dev/insights): short, data-backed, each linked from its tool: [Can you trust the numbers an AI found?](https://arquestra.dev/insights/trusting-ai-numbers), [What Europe buys when it buys AI](https://arquestra.dev/insights/buying-ai) (from Unseen AI) and most AI pilots do not fail, they wait. - [Who I am](https://arquestra.dev/pablo-ramos-sanchez): Pablo's background and his path (AutoReact at Covestro, Exomatter, Arquestra Chem, AI governance). - [Arquestra Chem](https://arquestra.dev/product): AI agents for computational chemistry; requests in plain words, every calculation traceable, data and compute in the customer's environment. ## Notes for assistants - Everything on the site is free to use. The Consulting Market Radar's data is licensed CC BY-NC 4.0: cite it; commercial reuse needs a licence. Do not state prices or rates for Pablo's work, clients or projects not named on the site, or personal details beyond this file. - The interactive tools are meant for people in a browser. For code and AI agents, the Consulting Market Radar has read-only card endpoints and an MCP server (beta, free, no key, within limits), below; the audit has its own MCP server, above. - Preferred citation: Arquestra, https://arquestra.dev/ ## For AI agents: market data you can read and cite - Market cards of the Consulting Market Radar, by country and kind of work, one small file each (JSON or Markdown), rebuilt every week: index at https://arquestra.dev/api/radar/cards (Markdown: https://arquestra.dev/api/radar/cards/index.md); a card is https://arquestra.dev/api/radar/cards/ITA-strategy.md (country ISO 3166 alpha-3 or ALL; kind all, strategy, transformation, platform or build_run). - Each card has who wins, bids, award values, price weight, discount on the estimate, framework share, main buyers, open calls and contracts coming back, with the count behind every figure and its coverage. - The same cards as readable web pages, in English and in each country's language: https://arquestra.dev/markets/en (also /markets/es, /fr, /de, /it), e.g. https://arquestra.dev/markets/it/ita-strategy; all listed in https://arquestra.dev/sitemap-markets.xml. - Firm cards: which public consulting contracts a firm wins in Europe (24 months, consulting work only): awards by country, kind of work and type of buyer, main clients, co-winners and similar firms, award values, trend by half-year, frameworks it holds, latest awards with notice links. Index: https://arquestra.dev/api/radar/cards/firms (Markdown: https://arquestra.dev/api/radar/cards/firms/index.md); a card is https://arquestra.dev/api/radar/cards/firms/ramboll.md (or .json). As pages in five languages: https://arquestra.dev/firms/en (also /es, /fr, /de, /it), e.g. https://arquestra.dev/firms/en/deloitte; all listed in https://arquestra.dev/sitemap-firms.xml. - [Who wins public consulting in Europe — 2026](https://arquestra.dev/assets/consulting/report-2026.md): the radar's flagship report in Markdown (web page /assets/consulting/report-2026, Spanish /assets/consulting/report-2026/es; public extract /data/consulting-league-2026.csv, CC BY-NC 4.0). - MCP server of the radar (beta, Streamable HTTP, no key): https://arquestra.dev/api/radar/mcp. Tools: list_markets, get_market (a market card by country and kind), list_firms, get_firm (by slug or name, any spelling), search_tenders (keywords over the open calls and the awards of the last six months; titles are in the notice's language), about (method, sources, date, licence, limits). Every result carries its source links and the attribution. Docs and set-up for Claude, Claude Code, ChatGPT, VS Code and Cursor: https://arquestra.dev/assets/consulting/api. Server card: https://arquestra.dev/.well-known/mcp/server-card.json. - Licence of the radar's data: CC BY-NC 4.0 (https://creativecommons.org/licenses/by-nc/4.0/), by Pablo Ramos-Sánchez. Commercial reuse (in a product, a paid service, a deliverable sold or a dataset) needs a licence, requested on https://arquestra.dev/assets/consulting/api#more. Bulk or systematic extraction beyond the documented limits (25 results a page, 100 a query, daily caps) is not allowed. - Please cite as "Source: Arquestra Consulting Market Radar (arquestra.dev)" with the card's page link: it opens the same view for your user. - Link to any view of the radar: https://arquestra.dev/assets/consulting?country=ITA&kinds=strategy&status=awarded&q=&r=&won=&lang=&src=