# Mickael Rutembya — full public context

> AI Systems · Data Engineering · Verifiable Infrastructure. Paris, France · onsite, hybrid or remote engagements.

Contact: mickael.pro.esn@proton.me
Website: https://mickael-umt.com/
LinkedIn: https://www.linkedin.com/in/mickael-r-57a9a8146/
GitHub: https://github.com/RickOwri
Calendar: https://cal.com/mickael-rutembya-eue7i8

## Operating boundary

The public website is served by Django/ASGI from a sovereign Hetzner deployment through Caddy, optionally fronted by Cloudflare. Runtime storage is local SQLite. No external API is required for pricing, qualification or persistence. The mission pricing calculation is a public, deterministic projection: every formula and coefficient is published so the exact amount can be reproduced offline. Only the promotional-code table, owner-only lead assessments and administrative data are excluded.

## Pricing calculation (public and reproducible)

The mission quote is a deterministic function of a few parameters. Every formula, constant and coefficient is published, so an LLM can recompute the exact amount offline, then verify it against the open compute endpoint. Promotional codes are the only pricing lever kept private (an unknown code is ignored).

### For AI agents (MCP tools — recommended)

Use these read-only MCP tools to estimate mission costs:

- **`get_pricing_spec`**: Fetch the complete specification with all formulas, coefficients, accepted input keys, worked examples and specialist day-rate cards.
- **`compute_mission_quote`**: Submit mission parameters and get a deterministic estimate including day rates, cost breakdown, totals, schedule recommendation and applied coefficients.

### For all (REST & pages)

- Interactive calculator: https://mickael-umt.com/outils/calculator/
- Methodology (explains every formula and design choice): https://mickael-umt.com/outils/calculator/methode/
- Full specification (JSON, formulas + coefficients + worked example): https://mickael-umt.com/api/public/pricing-spec
- REST endpoint (GET query string or POST JSON, deterministic, no storage): https://mickael-umt.com/api/public/quote

## Offers

### Agentic AI Engineering

- URL: https://mickael-umt.com/expertises/agentic-ai/
- Category: IT & data platforms (`it`)
- Promise: Turn promising agents into observable, controllable and operable services.
- Outcome: An agent architecture with explicit permissions, governed memory, evaluations, human oversight and execution evidence.
- Evidence: Internal R&D on multi-agent orchestration, execution consoles, review queues and verifiable runtime receipts.
- Stack: Python · FastAPI · MCP · RAG · evaluation harnesses · PostgreSQL
- Typical format: 3–12 months · 2–5 days/week · technical lead or embedded engineer

### Data Engineering & Evidence Architecture

- URL: https://mickael-umt.com/expertises/data-engineering/
- Category: IT & data platforms (`it`)
- Promise: Turn scattered data into traceable and reusable decisions.
- Outcome: Data models, ETL/ELT, events, quality controls, lineage, evidence stores and auditable APIs.
- Evidence: Architecture of data-intensive systems, append-only records, provenance and transactional invariants.
- Stack: Python · SQL · PostgreSQL · Pandas · PySpark · Airflow · Celery
- Typical format: 3–9 months · data backbone · architecture and implementation

### AI Security, Assurance & DevSecOps

- URL: https://mickael-umt.com/expertises/ai-security/
- Category: IT & data platforms (`it`)
- Promise: Make security a release condition, not a late report.
- Outcome: Quality gates covering code, agents, MCP, connectors, infrastructure, dependencies and smart contracts.
- Evidence: Design of Jenkins pipelines for AI security, software supply chain and release decision evidence.
- Stack: Jenkins · Terraform · Docker · Trivy · SBOM · policy gates · threat modelling
- Typical format: 3–6 months · 2–4 days/week · assurance and industrialisation

### Blockchain & Verifiable Computing

- URL: https://mickael-umt.com/expertises/blockchain/
- Category: IT & data platforms (`it`)
- Promise: Connect claims, decisions and executions to verifiable evidence.
- Outcome: Smart contracts, attestations, identity, account abstraction, multisignature governance and ZK/ZKML research.
- Evidence: Audited Lottery dApp, tokenized-voting backend, Rust Folder Mapper and Encode Club track with distinct provenance levels.
- Stack: Solidity · Rust · EVM · Solana · Foundry · Hardhat · Anchor · EZKL
- Typical format: 3–12 months · architecture, applied R&D and delivery

### AI Training & Adoption Engineering

- URL: https://mickael-umt.com/expertises/training/
- Category: Adoption & growth (`growth`)
- Promise: Turn training into observable skills that transfer to the workplace.
- Outcome: Objectives, exercises, rubrics, usage risks, transfer evidence and a measurable adoption programme.
- Evidence: Professional AI, Python, ML, cybersecurity and blockchain programmes since 2021.
- Stack: Python · AI adoption · cybersecurity · blockchain · assessment · document control
- Typical format: 3–6 month programme or focused sessions · onsite and remote

### Marketing Engineering & Context Systems

- URL: https://mickael-umt.com/expertises/marketing-engineering/
- Category: Adoption & growth (`growth`)
- Promise: Turn commercial context into a measurable, versioned and reusable system.
- Outcome: Avatars, offers, sources, prompts, analytics, content and decisions connected to an evidence register.
- Evidence: Controlled marketing workbook connecting mission, offers, evidence, avatars, messaging and conversion.
- Stack: Research · structured context · automation · analytics · prompt systems
- Typical format: 3–6 months · marketing operating system and automation

### Regulatory Technology & CSV Engineering

- URL: https://mickael-umt.com/expertises/reg-csv/
- Category: HealthTech & regulated environments (`healthtech`)
- Promise: Translate pharmaceutical risk into validation requirements, technical controls and defensible evidence before inspection or AI-workflow promotion.
- Outcome: IQ/OQ/PQ plans, GxP traceability, controlled sources and versions, acceptance tests, decision rights and validation packages connected to product risk.
- Evidence: Open-source PharmD studio: reproducible clinical workflows, ADMET benchmarks, FAERS signal detection and OHDSI integrations.
- Stack: Python · SQL · validation frameworks · audit automation · CDISC · OHDSI/OMOP
- Typical format: 3–12 months · freelance or consulting · regulated

### Clinical Data Management Automation

- URL: https://mickael-umt.com/expertises/cdm-automation/
- Category: HealthTech & regulated environments (`healthtech`)
- Promise: Reduce time-to-lock without obscuring the evidence chain across clinical data, automated rules, exceptions and human validation.
- Outcome: SDTM/ADaM pipelines, assisted query resolution, versioned quality controls, transformation lineage, exception queues and reconstructable lock decisions.
- Evidence: Clinical data pipelines, OMOP exports, Open Targets integrations and automated FAERS pipelines.
- Stack: Python · SQL · R · OpenClinica · Medidata Rave · CDISC · PySpark
- Typical format: 3–9 months · permanent/fixed-term · data backbone

### Pharmacovigilance Data Engineering

- URL: https://mickael-umt.com/expertises/pv-data-engineer/
- Category: HealthTech & regulated environments (`healthtech`)
- Promise: Accelerate signal detection and case processing without transferring safety judgement to a model or dashboard.
- Outcome: FAERS/openFDA pipelines, E2B(R3) NLP, traceable sources and versions, escalation thresholds, human-review queues and defensible signal decisions.
- Evidence: FAERS senolytic signal detection template, automated pipelines and Biolearn integration for RWE.
- Stack: Python · NLP · FAERS/openFDA · E2B(R3) · Biolearn · safety signal algorithms
- Typical format: 3–6 months · employed · PV operations

### HealthTech & PharmaTech Product Engineering

- URL: https://mickael-umt.com/expertises/healthtech-product/
- Category: HealthTech & regulated environments (`healthtech`)
- Promise: Turn pharma-client decisions into product features whose evidence, permissions, validation and ownership are designed with the product.
- Outcome: User stories connected to pharmaceutical risk, acceptance criteria, per-feature documentation, AI/data controls, human review and verifiable delivery.
- Evidence: PharmD R&D, OHDSI/OMOP mappings and Therapeutics Data Commons integrations aligned with business cases.
- Stack: Product management · Agile · CDISC · GxP · Veeva · Medidata · IQVIA · Cegedim
- Typical format: 6–12 months · vendor permanent · employed

## Ranked CV capabilities

Ranking is deterministic: minimum documented years, confidence, evidence strength and explicitly cited breadth. Each capability has a stable `/expertises/:id/` URL.

### #1 — Pharmaceutical and regulated operations (98/100)

- ID: `pharmaceutical-operations`
- URL: https://mickael-umt.com/expertises/pharmaceutical-operations/
- Family: Health & regulated environments
- Earliest defensible date: 2017
- Confidence: high
- Evidence: Employment, university records, hospital placements and quality control.
- Skills: Pharmacologie, Qualité, Radiopharmacie, Données cliniques, Confidentialité, Traçabilité

### #2 — Python and applied data processing (82/100)

- ID: `python`
- URL: https://mickael-umt.com/expertises/python/
- Family: AI & data engineering
- Earliest defensible date: 2021
- Confidence: high
- Evidence: Consulting activity, scripts, automation and dated teaching artifacts.
- Skills: Python, Pandas, NumPy, Automatisation, APIs, Traitement de données

### #3 — Technical training and instructional design (82/100)

- ID: `technical-training`
- URL: https://mickael-umt.com/expertises/technical-training/
- Family: Adoption & enablement
- Earliest defensible date: 2021
- Confidence: high
- Evidence: Contracts, programmes, materials, exercises, assessments and invoices.
- Skills: Instructional design, Évaluation, Python, IA, Cybersécurité, Blockchain

### #4 — Web and product engineering (67/100)

- ID: `web-engineering`
- URL: https://mickael-umt.com/expertises/web-engineering/
- Family: AI & data engineering
- Earliest defensible date: 2022
- Confidence: medium-high
- Evidence: Public repositories, deployed applications and client work.
- Skills: Django, React, JavaScript, TypeScript, HTML/CSS, Streamlit

### #5 — Solidity and EVM engineering (66/100)

- ID: `solidity-evm`
- URL: https://mickael-umt.com/expertises/solidity-evm/
- Family: Verifiable systems
- Earliest defensible date: 2023-05-30
- Confidence: high
- Evidence: Public Lottery dApp commits and Encode Club certificates.
- Skills: Solidity, EVM, Hardhat, Foundry, ERC-20, Tests de contrats

### #6 — Rust and systems tooling (65/100)

- ID: `rust`
- URL: https://mickael-umt.com/expertises/rust/
- Family: Verifiable systems
- Earliest defensible date: 2023-09-15
- Confidence: high
- Evidence: Public Folder Mapper repository and commit history.
- Skills: Rust, CLI, Filesystem, SQLite, Data tooling

### #7 — Data science and machine learning (63/100)

- ID: `data-science-ml`
- URL: https://mickael-umt.com/expertises/data-science-ml/
- Family: AI & data engineering
- Earliest defensible date: 2022
- Confidence: medium
- Evidence: Notebooks, teaching material, repositories and deliverables; the earliest notebook date remains to be consolidated.
- Skills: Scikit-learn, PyTorch, TensorFlow, BERT, Computer vision, Explainability

### #8 — LLM applications and RAG (55/100)

- ID: `llm-applications`
- URL: https://mickael-umt.com/expertises/llm-applications/
- Family: AI & data engineering
- Earliest defensible date: 2023
- Confidence: medium
- Evidence: LangChain/Streamlit prototypes, notebooks, client artifacts and teaching materials.
- Skills: LLM, RAG, LangChain, Prompting, Tool use, Evaluation

### #9 — Agentic AI systems (50/100)

- ID: `agentic-ai`
- URL: https://mickael-umt.com/expertises/agentic-ai/
- Family: AI & data engineering
- Earliest defensible date: 2024 / 2025 structuré
- Confidence: high
- Evidence: Agent registries, execution models, orchestration, review workflows and runtime specifications.
- Skills: Agent orchestration, MCP, Tool routing, Governed memory, Human review, Observability

### #10 — Data architecture, DDD and events (50/100)

- ID: `data-architecture-ddd`
- URL: https://mickael-umt.com/expertises/data-architecture-ddd/
- Family: AI & data engineering
- Earliest defensible date: 2024
- Confidence: high
- Evidence: Internal models, SQL backups, specifications and implementation plans under NDA.
- Skills: PostgreSQL, DDD, Event sourcing, Invariants, Schema evolution, Evidence stores

### #11 — AI governance and assurance (49/100)

- ID: `ai-governance`
- URL: https://mickael-umt.com/expertises/ai-governance/
- Family: Assurance & governance
- Earliest defensible date: 2024
- Confidence: high
- Evidence: Internal specifications, assessment frameworks, controlled training systems and governance documentation.
- Skills: Risk classification, Human oversight, Provenance, Data minimization, Audit readiness

### #12 — Jenkins, Terraform and supply chain (43/100)

- ID: `jenkins-terraform`
- URL: https://mickael-umt.com/expertises/jenkins-terraform/
- Family: Assurance & governance
- Earliest defensible date: 2025-2026
- Confidence: medium-high
- Evidence: Groovy pipelines, IaC files, architecture assessments and security configurations.
- Skills: Jenkins, Terraform, SBOM, Container scanning, Policy as code, Rollback

### #13 — Solana development (42/100)

- ID: `solana`
- URL: https://mickael-umt.com/expertises/solana/
- Family: Verifiable systems
- Earliest defensible date: 2023-10
- Confidence: to-confirm
- Evidence: Solana Bootcamp certificate and exercises; public repository evidence remains to be completed.
- Skills: Solana, Rust, Anchor, Account model, TypeScript

### #14 — AI system evaluation (41/100)

- ID: `ai-evaluation`
- URL: https://mickael-umt.com/expertises/ai-evaluation/
- Family: Assurance & governance
- Earliest defensible date: 2025
- Confidence: high
- Evidence: Harness specifications, scenario libraries, release criteria and consulting artifacts.
- Skills: Evaluation harnesses, Regression, Red teaming, Golden datasets, Release gates

### #15 — AI system security (41/100)

- ID: `ai-security`
- URL: https://mickael-umt.com/expertises/ai-security/
- Family: Assurance & governance
- Earliest defensible date: 2025
- Confidence: high
- Evidence: Threat models, evaluation plans and pipeline architectures, partly NDA-restricted.
- Skills: Threat modelling, Prompt injection, MCP security, Secrets, Agent permissions

### #16 — Zero knowledge and ZKML (34/100)

- ID: `zk-zkml`
- URL: https://mickael-umt.com/expertises/zk-zkml/
- Family: Verifiable systems
- Earliest defensible date: 2023-2024
- Confidence: to-confirm
- Evidence: ZK and zkML certificates, experiments and confidential applied research.
- Skills: ZK proofs, ZKML, EZKL, Attestations, Selective disclosure

## Queryable experience

- [Independent AI, Data and Software Engineering Consultant](https://mickael-umt.com/parcours/independent-ai-data-consulting/) — 2021–présent · Design of AI workflows, agent architectures, data systems, APIs, smart contracts and training programmes. · capabilities: python, data-science-ml, web-engineering, llm-applications, data-architecture-ddd, agentic-ai
- [Technical Instructor — Python, cybersecurity, data and blockchain](https://mickael-umt.com/parcours/technical-instructor/) — 2021–présent · Design and delivery of training with materials, exercises, assessments and practical scenarios. · capabilities: technical-training, python, data-science-ml, solidity-evm
- [Technical Specialist — privacy-preserving machine learning](https://mickael-umt.com/parcours/privacy-preserving-ml-research/) — 2024–présent · Decentralized science, applied cryptography and secure scientific and medical data sharing. · capabilities: zk-zkml, ai-governance, data-science-ml
- [R&D — governed AI operations](https://mickael-umt.com/parcours/governed-ai-operations-rd/) — 2025–présent · Agent registries, supervision, permissions, governed memory, human review and runtime evidence. · capabilities: agentic-ai, ai-evaluation, ai-security, jenkins-terraform
- [Community pharmacy — Strasbourg and Paris](https://mickael-umt.com/parcours/pharmacy-operations/) — 2015–2023 · Preparation, dispensing, advice, prescription analysis and Python automation of operational data. · capabilities: pharmaceutical-operations, python
- [Strasbourg University Hospital — radiopharmacy](https://mickael-umt.com/parcours/strasbourg-radiopharmacy/) — 2018–2019 · Radiochromatographic analysis, quality control, compliance and biological data processing. · capabilities: pharmaceutical-operations, data-science-ml
- [University of Strasbourg — university pharmacy curriculum](https://mickael-umt.com/parcours/pharmacy-studies-strasbourg/) — 2014–2020 · Six years of pharmacology, toxicology, pharmacokinetics, quality and hospital practice. · capabilities: pharmaceutical-operations

## Verifiable certificates

- [zkML Bootcamp](https://mickael-umt.com/parcours/certifications/zkml-bootcamp/) — mars 2024 – avr. 2024 · Encode Club · Encode Club · capabilities: zk-zkml, data-science-ml
- [ZK Bootcamp](https://mickael-umt.com/parcours/certifications/zk-bootcamp/) — juil. 2023 – sept. 2023 · Encode Club · Encode Club · capabilities: zk-zkml
- [Expert Solidity Bootcamp](https://mickael-umt.com/parcours/certifications/expert-solidity-bootcamp/) — juil. 2023 – sept. 2023 · Encode Club · Encode Club · capabilities: solidity-evm
- [Solana Bootcamp](https://mickael-umt.com/parcours/certifications/solana-bootcamp/) — sept. 2023 – oct. 2023 · Encode Club · Encode Club · capabilities: solana
- [Solidity Bootcamp](https://mickael-umt.com/parcours/certifications/solidity-bootcamp/) — avr. 2023 – juin 2023 · Encode Club · Encode Club · capabilities: solidity-evm
- [Token Engineering Fundamentals](https://mickael-umt.com/parcours/certifications/token-engineering-fundamentals/) — 2023 · Token Engineering Academy · Token Engineering Academy · capabilities: solidity-evm

## Public project evidence

### Lottery dApp (Mai–juin 2023)

- Summary: Bootcamp fork extensively developed by Mickael: backend, contract, tokens, closing, withdrawals, prize logic and debugging.
- Provenance: Verified substantial contribution
- Evidence: Audit: 26 of 29 commits attributed to RickOwri. It is never presented as the initial creation.
- Stack: Solidity · EVM · TypeScript
- Repository: https://github.com/RickOwri/Encode-Club-Solidity-Bootcamp-Project-W5-LotteryDapp
- History: https://github.com/RickOwri/Encode-Club-Solidity-Bootcamp-Project-W5-LotteryDapp/commits/main/

### Tokenized Voting Backend (Mai–juin 2023)

- Summary: NestJS/ethers backend on a scaffold: supply, balances, receipts, minting, transfer, delegation, voting power and Sepolia voting.
- Provenance: Verified training implementation
- Evidence: Four public commits, all attributed to RickOwri. Business logic is distinguished from the NestJS starter.
- Stack: NestJS · TypeScript · ethers.js · Sepolia
- Repository: https://github.com/RickOwri/Encode-Club-Solidity-Bootcamp-Project-W4-VotingDapp-Backend
- History: https://github.com/RickOwri/Encode-Club-Solidity-Bootcamp-Project-W4-VotingDapp-Backend/commits/main/

### Mistral Hackathon — Persona & Bias Evaluation (2024)

- Summary: Big Five/IPIP, JSON, Mistral and LangChain prototype exploring persona consistency, constrained answers and bias.
- Provenance: Owner-attested collaborative contribution
- Evidence: Contribution confirmed by Mickael and a public '(mick)' notebook. The aggregate commit prevents attributing the whole repository.
- Stack: Python · Pandas · Mistral · LangChain · JSON
- Repository: https://github.com/RickOwri/mistral-hackaton
- History: https://github.com/RickOwri/mistral-hackaton/commits/master/

### AI/RAG Experiment Portfolio (2023–2024)

- Summary: Public evidence registry containing a Solidity Code Tutor RAG prototype, a LangChain router and a Scrapy crawler.
- Provenance: Owner-attested original artifacts · flattened private history
- Evidence: 28 RickOwri commits. Artifact authorship is attested, but some private imports did not preserve their full history.
- Stack: Python · RAG · Chroma · LlamaIndex · LangChain · Scrapy
- Repository: https://github.com/RickOwri/projects-cv
- History: https://github.com/RickOwri/projects-cv/commits/main/

### Folder Mapper (Septembre–octobre 2023)

- Summary: Rust CLI crawling a filesystem and persisting filename/path mappings in SQLite.
- Provenance: Verified original creation
- Evidence: Non-fork repository; all 7 commits attributed to RickOwri.
- Stack: Rust · SQLite
- Repository: https://github.com/RickOwri/folder_mapper
- History: https://github.com/RickOwri/folder_mapper/commits/master/

### Streamlit LangChain Prototype (Avril–octobre 2023)

- Summary: Bounded LLM prototype testing a Streamlit interface, Docker packaging and GitLab CI delivery.
- Provenance: Verified original creation
- Evidence: Non-fork repository; all 7 commits attributed to RickOwri.
- Stack: Python · Streamlit · Docker · GitLab CI
- Repository: https://github.com/RickOwri/streamlit-lang
- History: https://github.com/RickOwri/streamlit-lang/commits/main/

### Governed AI Operations (2025–présent)

- Summary: Confidential R&D on agent registries, observability, permissions, memory, human review and controlled rollout.
- Provenance: Confidential R&D
- Evidence: Described at responsibility and architecture level; code and data are not public.
- Stack: Python · FastAPI · PostgreSQL · MCP · IAM
- Repository: 
- History: 

## Multi-source market snapshot

Verified 2026-07-30. Only sources publishing an actual Q1, median and Q3 feed the boxplots. London/UK contract day rates and New York employee wages are deliberately kept as separate distributions.

### Day rates for specialties close to my profile

- Digital marketing, London: Q1 EUR 332/day (EUR 41,5/hour), median EUR 456/day (EUR 57,0/hour), Q3 EUR 607/day (EUR 75,9/hour). Native observation: £284 · £390 · £519 / jour; 6 months to 15 April 2026; n=55; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/london/digital%20marketing.do).
- IT training, United Kingdom: Q1 EUR 435/day (EUR 54,4/hour), median EUR 600/day (EUR 75,0/hour), Q3 EUR 688/day (EUR 86,0/hour). Native observation: £372 · £513 · £588 / jour; 6 months to 28 July 2026; n=20; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/uk/it%20trainer.do).
- Data Engineer, London: Q1 EUR 526/day (EUR 65,8/hour), median EUR 614/day (EUR 76,8/hour), Q3 EUR 728/day (EUR 91,0/hour). Native observation: £450 · £525 · £623 / jour; 6 months to 26 July 2026; n=622; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/london/data%20engineer.do).
- Blockchain, United Kingdom: Q1 EUR 572/day (EUR 71,5/hour), median EUR 614/day (EUR 76,8/hour), Q3 EUR 702/day (EUR 87,8/hour). Native observation: £489 · £525 · £600 / jour; 6 months to 24 July 2026; n=17; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/uk/blockchain.do).
- Cybersecurity, London: Q1 EUR 562/day (EUR 70,3/hour), median EUR 661/day (EUR 82,6/hour), Q3 EUR 799/day (EUR 99,9/hour). Native observation: £481 · £565 · £683 / jour; 6 months to 18 July 2026; n=522; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/london/cyber%20security.do).
- AI / LLM, London: Q1 EUR 561/day (EUR 70,1/hour), median EUR 672/day (EUR 84,0/hour), Q3 EUR 818/day (EUR 102,3/hour). Native observation: £480 · £575 · £700 / jour; 6 months to 29 July 2026; n=1649; [IT Jobs Watch](https://www.itjobswatch.co.uk/contracts/london/artificial%20intelligence.do).


### New York tech wages, as hourly or daily equivalents

- Training & Development, New York–Newark–Jersey City: Q1 EUR 192/day (EUR 24,0/hour), median EUR 270/day (EUR 33,8/hour), Q3 EUR 361/day (EUR 45,1/hour). Native observation: $27.37 · $38.39 · $51.33 / heure; OEWS wages, May 2025; [BLS OEWS · CareerOneStop](https://cloudfront.careeronestop.org/Toolkit/StateAndLocal/Wages.aspx?dataview=table&location=New+York&soccode=131151).
- Marketing / Market Research, New York–Newark–Jersey City: Q1 EUR 258/day (EUR 32,3/hour), median EUR 339/day (EUR 42,4/hour), Q3 EUR 441/day (EUR 55,1/hour). Native observation: $36.78 · $48.33 · $62.84 / heure; OEWS wages, May 2025; [BLS OEWS · O*NET](https://www.onetonline.org/link/localwages/13-1161.00?st=NY).
- Data Scientist / AI, New York–Newark–Jersey City: Q1 EUR 342/day (EUR 42,8/hour), median EUR 459/day (EUR 57,4/hour), Q3 EUR 585/day (EUR 73,1/hour). Native observation: $48.71 · $65.37 · $83.25 / heure; OEWS wages, May 2025; [BLS OEWS · O*NET](https://www.onetonline.org/link/localwages/15-2051.00?st=NY).
- Information Security Analyst, New York–Newark–Jersey City: Q1 EUR 364/day (EUR 45,5/hour), median EUR 474/day (EUR 59,3/hour), Q3 EUR 593/day (EUR 74,1/hour). Native observation: $51.83 · $67.53 · $84.48 / heure; OEWS wages, May 2025; [BLS OEWS · O*NET](https://www.onetonline.org/link/localwages/15-1212.00?st=NY).
- Software / Blockchain, New York–Newark–Jersey City: Q1 EUR 441/day (EUR 55,1/hour), median EUR 563/day (EUR 70,4/hour), Q3 EUR 696/day (EUR 87,0/hour). Native observation: $62.81 · $80.21 · $99.02 / heure; OEWS wages, May 2025; [BLS OEWS · O*NET](https://www.onetonline.org/link/localwages/15-1252.00?st=NY).



### Bibliography and cross-checks

- IT Jobs Watch — London / UK · AI, Data, cyber, blockchain, training and marketing contracts (boxplot): Primary contract boxplot source: published Q1, median and Q3, with counts. [source](https://www.itjobswatch.co.uk/contracts/london/artificial%20intelligence.do).
- BLS OEWS · O*NET — New York and United States · occupational wages (boxplot): Primary salary boxplot source: May 2025 OEWS hourly percentiles. [source](https://www.onetonline.org/link/localwages/15-2051.00?st=NY).
- BLS OEWS · CareerOneStop — New York and United States · training and development (boxplot): Complements O*NET with 2025 percentiles for training roles. [source](https://cloudfront.careeronestop.org/Toolkit/StateAndLocal/Wages.aspx?dataview=table&location=New+York&soccode=131151).
- Malt — Paris · freelance Data Science and cybersecurity (crosscheck): Local cross-check: active-profile averages, not used as quartiles. [source](https://www.malt.fr/t/barometre-tarifs/expert-data/data-scientist).
- Free-Work — Île-de-France · Data Engineer day rates (crosscheck): Self-reported local cross-check; the 5–10 year average is not a percentile. [source](https://www.free-work.com/fr/tech-it/data-engineer/rate-tjm-freelance).
- Upwork — International USD market · AI and Machine Learning (crosscheck): Published ranges: convertible, but not relabelled as Q1/Q3. [source](https://www.upwork.com/hire/artificial-intelligence-engineers/cost/).
- Freelancer — International market · visible AI profiles (crosscheck): Small platform-ranked sample retained as exploratory evidence. [source](https://www.freelancer.com/hire/artificial-intelligence).
- Indeed — Paris · Data Scientist salaries (crosscheck): Average salary and reported volume: useful as a check, not as quartiles. [source](https://fr.indeed.com/career/data-scientist/salaries/Paris-%2875%29).
- Hellowork — Île-de-France · median Data Engineer salary (crosscheck): Local median without directly comparable Q1/Q3. [source](https://www.hellowork.com/fr-fr/salaires/data-engineer.html).
- YunoJuno — United Kingdom · 261,000+ freelance contracts (crosscheck): Aggregate UK-level cross-check across disciplines. [source](https://www.yunojuno.com/press-release/yunojuno-releases-2025-freelancer-rates-report-ai-skills-drive-shift-in-contingent-workforce-market-as-average-day-rates-rise).
- Banque centrale européenne — GBP/USD to EUR conversion (method): 27 July 2026 rates: EUR 1 = GBP 0.85524 = USD 1.1389. [source](https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/index.en.html).

GBP and USD observations are converted with the ECB reference rates dated 2026-07-27. One day equals 8 hours for unit switching only. Salary equivalents are not freelance TJMs. Malt, Free-Work, Upwork, Freelancer, Indeed, Hellowork and YunoJuno remain cross-checks and are not relabelled as quartiles.

## Legacy calculator references

These P25/median/P75 figures remain available for calculator compatibility. They are not the evidence source for the multi-source market snapshot above and are not automatic contractual prices.

- Pharmacovigilance / Data Engineer: P25 €340, median €420, P75 €500; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Clinical Data Management: P25 €380, median €460, P75 €540; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- HealthTech / PharmaTech Product Engineer: P25 €420, median €530, P75 €620; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Growth / Marketing Engineering: P25 €510, median €570, P75 €640; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Data Engineer: P25 €540, median €650, P75 €740; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Cybersecurity Consultant: P25 €630, median €750, P75 €880; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Blockchain Developer: P25 €610, median €750, P75 €860; source [TJMètre](https://tjmetre.fr/barometre/developpeur), verified 2026-07-22.
- AI / LLM: P25 €680, median €820, P75 €930; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- IT Trainer: P25 €500, median €820, P75 €910; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.
- Regulatory / CSV Engineer: P25 €750, median €1100, P75 €1350; source [TJMètre](https://tjmetre.fr/barometre), verified 2026-07-22.


## Agent access

### Model Context Protocol (MCP) — recommended for AI agents

The **MCP server** is the standard way for AI agents to read this site's structured data and call tools:

- **Endpoint**: `https://mickael-umt.com/mcp` (POST, JSON-RPC 2.0, Streamable HTTP)
- **Server registry**: `https://mickael-umt.com/mcp/server.json` (GET, automatic client configuration)

**Public read tools** (no authentication needed):
- `get_site_profile` — Identity and contact links
- `list_offers` — Consulting and engineering offers
- `list_capabilities` — Ranked CV capabilities (score, evidence, date, skills)
- `list_experience` — Career and training records with addressable proofs
- `list_certificates` — Verifiable certificates and related capabilities
- `list_projects` — Public projects with provenance and repositories
- `get_market_snapshot` — Multi-source tech market benchmarks (Q1/median/Q3)
- `get_public_rate_benchmarks` — Legacy calculator P25/median/P75 references
- `qualify_project` — Non-invasive route recommendation (read-only, no storage)
- `get_pricing_spec` — Complete public pricing specification (reproducible offline)
- `compute_mission_quote` — Deterministic mission quote (query string or JSON)

MCP is language-agnostic and has client libraries for Python, JavaScript, Go and more. See [modelcontextprotocol.io](https://modelcontextprotocol.io/) for integration guides.

### Alternative formats

- Public JSON context: https://mickael-umt.com/api/agent/context
- OpenAPI 3.1 contract: https://mickael-umt.com/api/openapi.json
- Submit a brief over REST: POST https://mickael-umt.com/api/agent/briefs

Read operations are public. Brief submission stores personal information locally and therefore requires explicit consent, a dedicated bearer token and an idempotency key. Never send credentials, confidential datasets, health data or personal data that is not necessary to evaluate the project.