Paul Boëssé

AI-powered automated LinkedIn prospecting SaaS platform — Next.js, NestJS, PostgreSQL, Claude AI, Stripe, Mistral AI

• Full functional definition: definition of the platform's functional architecture, detailed specification writing, backlog prioritization,

• Design and orchestration of the multi-agent AI architecture: multi-step workflow specification (invitations → personalized messages → follow-ups → human handover) integrating LLM personalization

• AI scoring model design: system of 11 purchase signal detectors (keywords, profile visitors, engagement) coupled with automated scoring to qualify leads

• Advanced prompting methods

HRIS SaaS platform for training administration + training marketplace — team of 5 — Nextjs, Typescript, Node.js, PostgreSQL, Mistral AI API

• Engineering leadership: Agile management of a cross-functional team of 5 people (devs, QA, designers): sprint planning, tech reviews, architecture choices, recruitment

• 12 product modules delivered in 18 months while maintaining technical quality (CI/CD, E2E tests, code reviews, API)

• Management of technically complex topics: external APIs, marketplace, invoicing, automations, AI components;

• Measurable impact: -90% admin time per case thanks to AI integration (Mistral, OCR) — definition, prioritization, and delivery with the team

• End-to-end AI workflow design: RAG pipelines, agents, prompt engineering, n8n orchestration

• Technical roadmap owner: trade-offs between feature velocity, technical debt, and compliance (Qualiopi, e-invoicing)

• Parallel projects: implementation of commercial tool suite (CRM: Pipedrive, Calling: Aircall, Lead: Apollo), setup of support, response to tenders, creation of a community of +1000 trainers

Design, management & development of an online software for post-stroke visual rehabilitation, created to assist practitioners in patient management, session planning, and execution of rehabilitation and assessment exercises (custom-developed exercises adhering to complex scientific protocols), including automatic analysis & results display. The software has been in use during sessions since September 2024 and is undergoing scientific validation for wider dissemination.

Involvement across the entire product cycle:

• Requirements gathering from practitioners;

• Facilitation of definition workshops;

• Formalization of functional requirements;

• Prioritization of evolutions;

• Translation of business needs & scientific protocols into product solutions;

• Coordination of design and development;

• Development (Postgresql/Node.js/Express/React/TypeScript)

• Deployment (Scalingo) and platform security in a high-requirement context;

• Compliance with HDS and GDPR standards ensured: sensitive data governance, security and confidentiality standards.

• Security audit & penetration testing conducted by external providers

The project required a strong ability to bridge the gap between business users, functional constraints, and technical implementation, in an environment involving sensitive data security and hosting.