From business discovery to hands-on AI-assisted development.
I bridge business strategy, product design and hands-on implementation. I translate client needs into functional specifications, design workflows and UX/UI, and use AI coding tools to generate, modify, test and debug working application code locally. I collaborate with a technical co-developer who manages infrastructure, security and production deployment. I remain accountable for client delivery and user adoption.
Hands-on AI-assisted software development (Codex, Claude Code, and previously Cursor)
A clear handoff. Shared accountability.
I own the client journey and hands-on application work. My technical co-developer owns infrastructure and production. We make delivery decisions and validate integrations together.
- Olga
- Discovery, workflow mapping, business rules and functional specifications. UX/UI, agent prompts, AI-assisted coding, code edits, local debugging and functional tests. Iteration, demos, training and user feedback.
- Technical co-developer
- Local technical foundation and stack setup where needed. Servers, infrastructure, security, staging, production deployment and operational monitoring.
- Together
- Delivery decisions, integration validation and handover.
Discover & scope
I audit real working processes, map workflows and identify where time is lost or errors occur. I agree the first useful scope with the client.
Output: prioritised business needs and a scoped first module.
Specify & design
I turn discovery into functional specifications and business rules. I design user journeys, UX/UI, navigation and mockups before building modules.
Output: functional requirements and a usable product flow.
Build with AI coding tools
On the local technical foundation prepared by my co-developer, I use Codex and Claude Code to generate and modify application code, module by module. Earlier work used Cursor.
Environment: local Docker, Supabase and Mailpit; reusable skills support implementation.
Test/debug locally
I run the application locally, debug code, perform functional tests and iterate on UX/UI and business rules. I commit changes and prepare them for staging validation.
Output: working local modules, tested workflows and iterative fixes.
Production handoff
My co-developer deploys staging; I test client workflows and iterate on fixes. Together we validate integrations and delivery readiness. My co-developer handles production deployment, security and operational monitoring.
Handoff: local application → staging validation → production release by my co-developer.
Train & improve
I demonstrate the solution, train users and gather feedback. Client conversations, WhatsApp groups where appropriate and usage logs help me identify friction and prioritise improvements.
Focus: adoption, reliability, time saved and errors avoided.
Understand & choose
I assess the business need and decide whether an agent fits the client better than a structured application.
Output: a solution choice grounded in real working habits.
Specify the behaviour
I translate the client’s methods, vocabulary and decision criteria into functional specifications and agent instructions.
Output: agent responsibilities, business rules and human checkpoints.
Configure & connect
My co-developer installs Hermes Agent, configures instructions, memory and connectors, and connects the client’s working channels.
Environment: Hermes, a second brain, and channels such as Slack, Telegram or Discord.
Train & adopt
I take over user training and help the client incorporate the agent into everyday work. Technical configuration remains with my co-developer.
Handoff: a configured system → supported, trained users.
Three contexts. The same ownership.
An AI-agent deployment first, followed by a live business application and a field-tested SaaS pilot. Each case separates my contribution, technical collaboration and results.
Groupe Cayambe
Engineering & strategy consultancyI specified eligibility, GO/NO-GO criteria, prompts and orchestration for a live tender-agent system.
Agent deployment live02ETL
Accounting practiceI built and tested the local application with Codex; my co-developer deployed it to production.
Live deployment03ISG Paris
Business schoolA 2.5-week pilot build, tested during the December 2025 recruitment campaign.
Pilot delivered & field-testedTools support the method.
I use Codex and Claude Code to build and debug application code. In client solutions, Hermes-based agents execute the business workflows I specify. Reusable skills support requirements, UX/UI and implementation.
- My coding tools
- Codex · Claude Code · previously Cursor
- Build support
- Reusable skills for specifications, UX/UI and implementation
- Application stack
- React / Next.js · local Docker / Supabase / Mailpit
- Client agents
- Hermes Agent · specialist agents · second brain
- Technical co-developer
- Self-hosted Supabase · private GitHub · Coolify · OVHcloud · security monitoring
Application for Client Deployment Principal, AI (EMEA) · Valis
Continue with Groupe Cayambe