AI & Solidity engineering.
From code to production.
I'm Adam Boudjemaa, an AI engineer and Solidity developer, former CTO and named co-author of ERC-3643. Based in France, I build production agents, RAG systems and smart contracts. Available for freelance projects, fractional leadership and full-time roles.
Explore technical capabilities
// WHAT I BUILD
I work across the whole stack, so you are not hiring three specialists and an integrator. The engagement models further down are just the commercial shape this takes.
AI agents & orchestration
Multi-agent systems, tool use and function calling, planning and routing, human-in-the-loop approval, least-privilege agent permissions, cost and latency budgets.
RAG & data
Retrieval pipelines, chunking and embedding strategy, hybrid and re-ranked search, source-grounded answers with verified citations, abstention when the context does not support an answer.
Evaluation & reliability
Eval suites that gate a release, regression sets, LLM-as-judge with human calibration, tracing and observability, incident review, drift monitoring.
Backend & platform
APIs and services, event-driven pipelines, queues and workers, Postgres and vector stores, cloud infrastructure, CI/CD, containerisation, secrets and access control.
Blockchain & protocol
Smart contract design and review, token standards, permissioned and compliance-aware transfers, identity and claims, account abstraction, audit preparation.
Regulated delivery
Traceable decisions, human approval gates, data access controls, model and vendor evaluation, and documentation for your legal and compliance reviewers.
Frontend & product
React and Next.js applications, design systems, accessibility, dashboards and internal tools, performance work.
Leadership & enablement
Technical strategy, architecture decision records, hiring and interview loops, code review culture, mentoring, and handing the system over so your team owns it.
// ENGAGEMENT MODELS
Architecture Review
A technical review of your AI system, backend or smart contracts. You receive a written report with findings, recommended fixes and priorities based on risk and effort.
- Architecture, data flow and failure modes
- Findings ranked by risk and effort
- Written report with the exact fixes
- 1 follow-up call
- Timeline agreed after scoping
AI Agent Systems
An end-to-end build: retrieval, agents, APIs, data, release evaluations and the interface your team uses. On a previous fund engagement, I measured a bank-reconciliation workflow going from 8 hours to 23 minutes.
- Design and build, front to back
- Retrieval, agents, API, data and UI
- Eval suite and observability included
- Milestone-based delivery
- Slack access during build
- Knowledge transfer + docs so you own it
Fractional CTO
Hands-on technical leadership: technical strategy, architecture, hiring, vendor and security decisions, and shipping alongside your engineers. Former CTO at Integra, where I built the engineering team and led protocol delivery. Scope and commitment agreed before we start.
- Time commitment agreed with your team
- Strategy, architecture and roadmap
- Hiring, interview loops and mentoring
- Hands-on in the codebase, not just review
- Contract terms agreed upfront
- Direct Slack/async access
// SPECIALIZED SERVICES
ERC-3643 Implementation Audit
$5,000-10,000I'm a named author of ERC-3643. I review your implementation against the standard: identity registry, required claims, compliance modules and transfer restrictions. You receive documented findings and recommended fixes within an agreed scope.
▸ DISCUSS THE SCOPERWA Tokenization Advisory
$2,500/dayTechnical advisory for teams building tokenized real-world assets. Standard selection (ERC-3643 vs ERC-1400 vs custom), compliance architecture, cross-jurisdictional identity, institutional-grade smart contract design.
▸ DISCUSS THE SCOPEWhen scoping an RWA project, decide where the assets should live. That argument, and why a general-purpose chain keeps failing regulated assets, is written up in why RWA tokenization needs its own blockchain. The standard-selection question that follows it is answered on ERC-6960, for fractional ownership of a single asset, and on ERC-7410, for revoking a spender allowance.
// RESULTS
Verified operating record
Distinct systems and attributable results
11
Agent platform for a confidential fund
First-party: delivered on the PE-fund engagement
~46
Agents for Integra internal operations
First-party: from Integra internal operations, where I was CTO
8h → 23min
Bank reconciliation, automated
First-party: measured before and after on the PE-fund engagement
$441M
On-chain volume processed
First-party: my record of the Biconomy infrastructure I led
60%
Verified Integra AWS cost reduction
First-party: Integra billing, $5,506/mo down to $2,186/mo
// HOW IT WORKS
First Conversation
Share the problem or role, the team and the target date. We establish whether my experience matches what you need.
Scope & Proposal
Agree the scope, timeline, commercial terms and how changes will be handled before work starts.
Build
I work in your codebase, share progress and review deliverables with your team at agreed milestones.
Handoff
Hand over the code, documentation and operating notes. Agree who owns maintenance and any follow-up support.
// SPEAKING & WORKSHOPS
No slides full of buzzwords. No "10 predictions for 2027." Real architecture decisions, real failure stories, real code. From intimate workshops to main stages. In English or French.
// TALK TOPICS
Multi-Agent Systems That Actually Ship
A field report from two distinct systems: an 11-agent platform delivered to production for a confidential real-estate fund, and approximately 46 agents deployed for Integra's internal operations. Coordination, tool design, evals, failure modes and operational handover.
AI Agents and Smart-Account Permissions
How an application can let an AI agent request on-chain actions while smart contracts enforce permissions. Account abstraction, bounded authority and the checks needed between model output and execution.
From ERC Draft to Ethereum Standard
Lessons from co-authoring Ethereum standards: defining interfaces, implementing the proposal and responding to review. ERC-3643 is Final; ERC-6960 and ERC-7410 are Drafts in the registry; ERC-8203 is a draft still under review.
Building an AI CTO: Automating Engineering Operations
Lessons from building AI tooling for engineering operations: task routing, code review, documentation and approval gates. How to decide what agents can own and where engineers retain control.
42 to CTO: Learning Through Projects
From the project-based program at 42 Paris to CTO of a Layer-1 protocol. A talk about learning through code, working with peers and taking on engineering responsibility. Available in French and English.
// FORMATS
Keynote
30-45 min main stage talk with Q&A
Workshop
Half-day hands-on session, 15-30 people
Podcast
Long-form conversation, remote or in-person
Not a fit?
If you need a WordPress site or a ChatGPT wrapper, I'm not your guy. I build production AI systems for companies where manual processes cost real money. If that's not your situation, I'll point you to someone who can help.
// FAQ
What's included in the ERC-3643 audit?
A scoped review of token behavior, identity, required claims, compliance modules and transfers against ERC-3643. You receive documented findings and recommended changes. I'm a named author of the standard. The scope must separately identify independent security review and the legal requirements supplied by your team.
What's your availability?
Current availability: AVAILABLE NOW. Send the role or project scope and your target start date so we can agree a workable schedule.
Do you write code or just advise?
Both. I write production code and can take responsibility for architecture, technical reviews and team delivery within the agreed scope.
What tech stack?
Solidity, TypeScript, Go, Python. Multi-agent AI systems (Claude, GPT, open-source). EVM chains. React and Next.js interfaces, APIs and data pipelines when the product needs them.
Can you speak at my event?
Yes. Keynotes, workshops and podcasts, in English or French. All talks can be customized for your audience. I prep properly. I don't phone it in.
What if it doesn't work out?
We agree milestones, review points and termination terms before starting. If the scope or collaboration needs to change, we review it together.
Technical questions instead, on ERC standards or on running AI agents in regulated finance? Those are answered on the FAQ.
Tell me the problem.
A project, a permanent role or a talk. Share the context and what you need me to own.
▶ START A CONVERSATIONPrefer email? contact@adam-boudjemaa.com