Christophe Bellec, independent architect
Software, Cloud & AI Architecture
I help companies design, audit and evolve their software systems: frontend, backend, AWS and AI integration.
Focused engagements of a few days, fixed-scope work, or regular ongoing support.
15 years
of development, architecture and production delivery
Software, cloud, industrial
from the web to systems that drive hardware
France and international
teams spread across France, the United States and India
Design and build
I recommend the architecture and I can implement it
Three domains, one discipline
Designing systems that hold up in production. Software, cloud and AI are three faces of that work, not three separate jobs.
Software
Software architecture
Understanding an existing system, deciding its structure, and writing the code that puts it in place, from the frontend through to the APIs.
- An architecture people can read, where everyone knows where a new feature belongs
- Technical debt scoped and prioritised rather than endured
- Stable APIs and contracts between teams and services
- TypeScript
- Node.js
- NestJS
- React
- Next.js
- PostgreSQL
Cloud
AWS architecture
Sizing, securing and industrialising infrastructure without making it more complex than it needs to be.
- Repeatable deployments and environments that match
- Costs understood, measured and reduced where possible
- Observability that lets you diagnose before the customer calls
- AWS
- ECS / Fargate
- RDS
- Redis
- CI/CD
- Observabilité
AI
AI integration and agents
Connecting LLMs, RAG and agents to your applications, data and business tools, with the guardrails a real production system requires.
- Use cases chosen for their value, not for the demo effect
- Integration wired into your systems, your permissions and your data
- Cost, traceability and monitoring handled from the design stage
- LLM
- RAG
- Agents
- MCP
- Tool calling
Focused engagements
Not every technical question needs six months of consulting. Many can be scoped, settled or prototyped in a few days.
Architecture audit
1 to 2 daysA read of the architecture, technical debt, risks and performance pressure points, followed by a debrief with a prioritised action plan.
Deliverable: System map, ranked risks and a sequenced action plan.
AWS and FinOps audit
1 to 2 daysA review of the cloud architecture, sizing, cost, security and observability, with the quick wins identified.
Deliverable: Costed diagnosis and a list of quick wins by impact.
AI opportunity audit
1 dayAn analysis of processes, systems and data to identify which AI use cases are genuinely worth pursuing, and whether they are feasible.
Deliverable: Prioritised use cases, feasibility and technical prerequisites.
AI architecture review
1 to 2 daysAn audit of an existing LLM, RAG or agent architecture: answer quality, security, cost, observability and readiness for production.
Deliverable: Breaking points identified and a route to production.
AI integration POC
3 to 5 daysA working prototype connected to a real application, documents, APIs or business tools, so you can decide on evidence.
Deliverable: A runnable prototype and a target architecture note.
AI visibility audit (GEO)
1 to 2 daysWhat AI assistants can actually extract from your site: content present in the initial HTML, an identifiable entity, statements they can reuse as-is, consistency with your public traces.
Deliverable: Extractability diagnosis and entity architecture.
Need something longer? An engagement can continue into development, architecture or technical leadership.
See all servicesWhat you get
Focused scoping, actionable recommendations, then hands-on support through implementation, without over-engineering. Your business priorities stay yours.
Architectural clarity
Documented decisions, technical debt scoped, a roadmap you can actually follow.
Risk under control
Security, infrastructure costs, dependencies and failure modes designed in from the start.
Pragmatic delivery
From diagnosis to production code: prioritisation, prototypes and maintainable implementation.
Knowledge transfer
Standards, code reviews and hands-on support so your team stays autonomous.
They trusted me
See the references
Recent writing
A few in-depth articles on the subjects I work on in the field.
How to connect an LLM to internal company data without reindexing everything
In most cases you reindex nothing. You let the model call the systems that already hold the data (through the existing application layer, with its permissions) and you only build a vector index for unstructured content that no API can query. Wholesale reindexing is a default decision, rarely a motivated one.
RAG or fine-tuning: the question is almost never framed correctly
RAG gives the model knowledge it does not have: your documents, your data, your context. Fine-tuning changes its behaviour: output format, tone, adherence to a taxonomy, one very repetitive task. A knowledge problem is not solved by fine-tuning, and a format problem is not solved by RAG. In practice, more than nine projects out of ten are a RAG problem or simply a prompting problem.
What AI integration really costs in a mid-sized company: the lines nobody quotes
The cost of an AI integration project splits into three blocks: getting the data and the surrounding systems into shape, often the heaviest; building the integration itself; and running it, which includes a per-use variable cost. Model calls are usually a minority share of the total. A proof of concept costing a few days says nothing about the cost of going to production, which is a software project in its own right.
Frequently asked questions
Do you only take on large projects?
No. Part of what I do is deliberately short: a one-day audit, an architecture review, a scoping exercise or a proof of concept over a few days. I can then support the implementation if that helps.
Can you add AI to an existing application?
Yes, that is exactly one of the main use cases: connecting an LLM, a RAG pipeline or agents to the APIs, data and business tools you already run, without starting over.
Do you work with OpenAI, Claude or another provider?
It depends on the need: the nature of your data, confidentiality constraints, latency, cost and existing architecture. I design integrations so the provider stays replaceable rather than locking you in.
Do you only advise?
No. I can audit, design, then implement the components that matter. That is what keeps my recommendations applicable.
Do you work remotely?
Yes. Most engagements run remotely, with on-site sessions in the Paris area when the project calls for it.
Do you work in English?
Yes, fluently (C1). My current engagement involves teams in France, the United States and India: specifications, code reviews and workshops all happen in English.
Paris area and remote
Based in the Paris area, I work on-site across Île-de-France and remotely throughout France. One person to talk to, direct and technical, with no middlemen.
Who is Christophe Bellec?A 30-minute call to scope your need
Describe your context and I’ll come back with a clear proposal.
Let’s talk about your project