Herve Roussel

Built an AI-assisted contract review platform that transforms uploaded legal documents into structured clause-by-clause analysis workflows. Users can upload PDF/DOCX contracts, receive plain-English explanations, risk scoring, AI-generated rewrite suggestions, and track inline edits with version history. I designed and implemented the full-stack architecture, including document ingestion, clause extraction, AI review pipelines, rich-text editing workflows, authentication, storage, and serverless back-end services. The system focuses on workflow usability and explainability rather than generic legal chat interfaces.

Built a full-stack AI research workflow that helps investment and strategy teams turn fragmented market signals into structured intelligence reports. Users can define company priorities, launch AI research agents, and generate evidence-backed newsletter-style reports with attribution and quality scoring. I designed and implemented the end-to-end architecture, including AI workflows, background orchestration, Supabase/Postgres data modeling, LLM integrations, observability, and the operator-facing UI. The platform combines AI-assisted extraction, semantic workflows, and structured reporting rather than generic chatbot interactions.

Built an AI-assisted due diligence platform that helps analysts and consulting teams transform public documents, PDFs, and web sources into source-grounded company research reports. The platform supports semantic search, cited Q&A, automated report generation, and multi-stage AI research workflows. I designed and implemented the full-stack system, including retrieval pipelines, vector search architecture, asynchronous job orchestration, OCR/document ingestion, AI synthesis workflows, and the research-focused user experience. The platform combines AI automation with structured workflows to accelerate business and investment research.

Built a full-stack workflow capture platform that records live browser interactions and converts them into reusable guided product tours. The system combines a Chrome extension, web dashboard, back-end APIs, and cloud storage to help teams document, share, and replay operational workflows inside real web applications. Users can record clicks and text inputs, automatically capture screenshots and UI targets, edit step instructions, and replay tours with contextual guidance overlays. I designed and implemented the end-to-end architecture, including browser extension logic, DOM instrumentation, front-end UX, authentication, back-end services, database modeling, and cloud deployment. The platform demonstrates workflow automation and operational tooling beyond simple documentation or screen recordings.

Built an AI-assisted traffic operations platform that converts live video feeds into structured monitoring and alert workflows. The system performs real-time vehicle detection, congestion monitoring, throughput analysis, and incident alerting while providing operators with review and dispatch workflows. I designed and implemented the architecture spanning computer vision services, real-time dashboards, WebSocket infrastructure, AI-assisted frame analysis, and workflow state management. The platform demonstrates how AI and computer vision can be integrated into operational decision-support systems.

Built a full-stack AI coaching prototype that helps users practice high-stakes conversations through live voice simulations with AI personas. Users select a scenario, engage in a real-time voice conversation powered by ElevenLabs Conversational AI, and receive structured coaching feedback afterward based on transcript analysis.

I designed and built the end-to-end product workflow, including front-end UX, voice AI integration, serverless back-end functions, transcript evaluation pipelines, and rubric-based scoring logic. The system evaluates communication quality across dimensions like empathy, curiosity, clarity, and engagement balance, then generates actionable coaching feedback and key conversational moments.

The project demonstrates how AI can be embedded into a practical training workflow rather than a generic chatbot experience. It also showcases rapid MVP development, AI workflow orchestration, prompt engineering, and full-stack product prototyping for commercially relevant SaaS applications.