Contact
SaaS platform for mortgage credit brokers powered by AI

SaaS platform for mortgage credit brokers powered by AI

A SaaS platform for mortgage brokers, automating workflows from loan case management to bank submission with AI.

2025 - 2026
CTO & Co-Founder (ACCENSEO)
Next.js 16React 19TypeScript 5.9Prisma 7PostgreSQL 16Tailwind CSS 4TerraformDockerTraefikVitest

Lines of Code

211,403

TypeScript (666 .tsx + 580 .ts)

Applications

7

7 Next.js 16 apps

DB Models

98

15 modular Prisma files

Planned Integrations

96

Stripe, Resend, Neon...

Presentation

A digital revolution for mortgage brokerage

This platform is the startup I co-founded as CTO and developer with an entrepreneur from the mortgage brokerage industry. Our ambition is clear: revolutionize the mortgage broker profession in France through a novel, AI-powered SaaS.

I designed and built the whole platform myself - a complete digital toolkit bringing together loan case management, financial simulations, CRM and compliance tracking - powered by artificial intelligence to automate repetitive tasks such as document data extraction or borrower information collection.

I built it as a Turborepo monorepo with 7 distinct Next.js 16 applications and 14 shared packages, replacing the manual, paper-based processes that still dominate the brokerage industry.

CTO & Co-FounderACCENSEOB2BMortgage brokers & insurance prosB2CBorrowers (client portal)
Project Nature

Multi-application SaaS platform (monorepo) - 7 distinct web applications + 14 shared packages, covering the entire mortgage brokerage workflow.

Business Domain

Finance / Real Estate - mortgage brokerage, borrower insurance, credit consolidation, professional loans. Compliance with ORIAS, GDPR, and HCSF usury rate regulations.

Platform Scope

Broker workspace

Dashboard, full loan case lifecycle (create, track, archive), document management, messaging, calendar

AI-powered borrower portal

"Zero Entry" mode - auto-extract data from documents via multimodal AI

Simulation suite

Borrowing capacity, installments, PTZ, notary fees, insurance, loan smoothing

Commercial CRM

Sales pipeline, prospect management, contracts, demo booking

SaaS admin dashboard

Client management, subscriptions, billing, feature flags, AI supervision, RBAC, audit logs, GDPR

Centralized auth hub

SSO, MFA (TOTP, WebAuthn, backup codes), magic link, phone auth

Marketing site & CMS

Product pages, pricing, blog (Payload CMS), case studies, integrations

Technical documentation

MDX site with Contentlayer, search, navigation, Mermaid diagrams

Design system

108 UI components, 6 layouts, 16 AI components, interactive previews

Infrastructure as Code

Terraform (5 modules, 3 environments), Docker Compose (5 services), 15 CI/CD workflows

Platform Architecture
7 Next.js 16 applications connected through shared packages

Objectives, Context, Stakes & Risks

Understanding the strategic vision behind the code

Objectives
  • Deliver an all-in-one solution for brokers: CRM, simulation, case management, compliance, banking integrations
  • Automate borrower data collection through multimodal AI (document extraction, conversational chat)
  • Ensure regulatory compliance (ORIAS, GDPR, HCSF usury rates)
  • Offer a multi-tenant SaaS model with white-labeling (custom domain, branding, SMTP, email templates)
  • Address the digital transformation of the French mortgage brokerage market
Stakes
  • - Position as the reference digital solution for French brokers
  • - Strict compliance with French financial regulations (usury rates, HCSF, ORIAS) and GDPR
  • - AI supervision and cost control with monitoring and versioned prompts
Identified Risks

AI costs

4 AI providers, strict cost monitoring required

ORIAS & ACPR compliance

Verify each broker ORIAS registration and display mandatory legal notices

Usury rate & HCSF

Respect quarterly usury rate and HCSF 35% debt-to-income cap

DCI & pre-contractual duties

Mandatory ESIS, written duty to advise, AML/KYC per loan case

Document retention

AML/KYC: 5 years, accounting: 10 years, ESIS: loan + 5 years

GDPR & financial data

Mandatory DPIA, encryption, RBAC, access and erasure rights

EU hosting & sovereignty

EU/EEA hosting required, French hosting recommended by ACPR

LCB-FT / KYC & Tracfin

Identity check, PEP detection, ongoing vigilance, Tracfin reporting

Audit trail & ACPR

Timestamped logs of every action, kept 5 years minimum

The Steps - What We Do

A concrete, phase-by-phase journey through the build

Four phases, each with my actions, my choices and the difficulty I had to beat.

Phase 1
Foundations & architecture
Pre-git (~2025)
  • Monorepo + stack (Next.js 16, Prisma 7, Tailwind v4, Better Auth)
  • 98-model Prisma schema across 15 modular files
  • 7 apps + 14 shared packages, each shippable alone
  • Difficulty: "schema and auth before pixels" discipline
Phase 2
Migration & restructuring
4 Feb 2026
  • Full monorepo upgrade to Next.js 16 (middleware → proxy)
  • Centralized auth hub, no duplicated login anywhere
  • Admin dashboard + security layer (rate limit, audit, HIBP)
  • Difficulty: stack upgrade on a 98-table DB, no migration tool
Phase 3
Feature enrichment
5-10 Feb 2026
  • MortgageApplication → LoanCase (broker vocabulary)
  • AI supervision: cost tracking, versioned prompts, human-in-loop
  • CRM pipeline (prospects, deals, contracts, demos)
  • Difficulty: TypeScript strict on 7 apps, batched refactors
Phase 4
Deployment & debugging
23 Feb 2026
  • Fixed Prisma bundling in Vercel production build
  • End-to-end SSO debug (secrets, cookies, callbacks, redirects)
  • Temporary debug endpoints, removed once root cause found
  • Difficulty: prod-only bugs → reproducible preview per branch

The Actors - Interactions

ACCENSEO team in co-founding with a brokerage client

The platform is a two-founder startup. I carry the whole technical side as CTO and sole developer. My business co-founder brings 15+ years of field expertise in mortgage brokerage - he knows the daily pains of brokers, the regulatory constraints, the banking partners, and the real-world workflow of a loan case from first client call to bank acceptance.

This duo is the backbone of the venture: one product vision grounded in deep industry knowledge, one technical architecture capable of delivering it.

My role (CTO & developer)
  • - Full-stack: architecture, frontend, backend, database, DevOps, CI/CD
  • - Product design, UX flows, design system, i18n and accessibility
  • - AI integration: Anthropic, OpenAI, Google, prompts, cost supervision, human-in-the-loop
  • - Infrastructure: Terraform, Docker, Vercel, Neon, Cloudflare R2, Traefik
  • - Leveraging AI coding assistants as productivity tools under my full review and architecture
Business co-founder (broker)
  • - Mortgage brokerage domain expertise (loan cases, banking, insurance)
  • - Regulatory knowledge: ORIAS, ACPR, HCSF, usury rates, AML/KYC
  • - Commercial network: existing banking partners and broker peers
  • - Product validation: real-world test bench for every feature we ship
  • - Go-to-market and first commercial customers
How we work together

Our ritual is lightweight and honest: a weekly product sync where he walks me through a real brokerage scenario, I demo what was shipped, and we prioritize the next iteration.

Between syncs, we use a shared backlog with strict "is this a broker pain?" filter - anything that does not pass gets dropped. Field feedback is treated as the single source of truth.

This direct loop between technical execution and domain expertise is what avoids the classic trap of engineers building features nobody needs.

Relational challenges overcome
  • - Bridging vocabularies: what a broker calls a "dossier" maps to a LoanCase with 10+ linked entities. Early in the project, we invested time in a shared glossary before writing any schema.
  • - Scope discipline: the regulated scope (ORIAS, HCSF, AML) is massive. We trained ourselves to say "not MVP" even on features that feel obvious, to avoid diluting time-to-market.
  • - Tempo alignment: a single developer moves in days, a commercial network in weeks. We built a two-track cadence - fast technical iterations and slower go-to-market steps that do not block each other.

Results

Measurable impact for the project and personal growth

Personal Growth
  • Deepening expertise in financial domain modeling (mortgage regulations, HCSF, usury rates)
  • Strengthening skills in centralized authentication patterns (SSO hub with MFA, WebAuthn, RBAC)
  • Building multi-app monorepo architecture at scale (7 apps, 14 packages)
  • Hands-on experience with multi-provider AI integration (Mistral, Anthropic, OpenAI, Google) and cost supervision
  • Developing Infrastructure as Code proficiency with Terraform multi-environment deployments
Effort Distribution Across 7 Applications
For the startup - measurable impact

The startup is pre-commercial launch. The measurable impact here is what was produced solo, under co-founder validation, in the time invested. Every line below is verifiable in the codebase.

  • Solo engineering delivery → 211,403 lines of TypeScript, 1,246 source files, 172 pages shipped as a single cohesive product
  • Monorepo architecture decision → 7 Next.js 16 applications and 14 shared packages produced by one developer, instead of a 4-5 engineer squad normally required for the same scope
  • Unified SSO hub → one authentication surface for 5 distinct user roles (admin, broker, client, docs, design), eliminating duplicated auth code in every app
  • AI-first workflows → "Zero Entry" mode removes 80% of manual data entry for borrowers by extracting data from their documents via multimodal AI
  • Regulated-scope schema → 98 Prisma models covering loan cases, usury rates, HCSF rules, ORIAS compliance and audit logs, ready for ACPR scrutiny
  • Integration platform → 96 integrations planned across banks, CRMs and productivity tools, positioning the product as a central hub in the French brokerage ecosystem
  • Infrastructure as Code → 3 reproducible environments (dev, staging, production) via Terraform, reducing environment drift risk to near-zero

Project Roadmap

What we are building next

Immediate future

Finalize SSO in production, onboard the first pilot brokerage cabinet with my co-founder, and start a controlled test of the AI-powered client portal on real loan cases. The database schema is already dimensioned to absorb complex banking data - usury rates, rate grids, HCSF rules, and rate history.

Medium-term evolution

Activate the planned banking integrations through open banking and document submission APIs, strengthen test coverage on critical paths (auth, billing, loan cases), break the single init migration into incremental migrations, and progressively bring in a small team around me to leave sole-developer territory.

Today

Today, the platform is in active pre-commercial development. The technical foundation has held up remarkably well: the monorepo has absorbed a complete pivot (domain migration, Next.js 16 upgrade, auth centralization) without breaking, which is already a lesson in itself about the durability of early architectural choices. The co-founding pair is stable and the roadmap toward commercial launch is clear. What remains to prove is operational: real users on real loan cases, not the code itself.

Critical Reflection

Honest retrospective on strengths, weaknesses and lessons learned

Strengths

  • AI as product differentiator

    "Zero Entry" mode + AI supervision position the product on an innovative segment.

  • Multi-layer fintech security

    MFA + SSO hub, AES-256 encryption, RBAC, Upstash rate limiting, immutable audit logs.

  • Exemplary monorepo architecture

    7 apps + 14 shared packages: code reuse, clear domain boundaries.

  • Complete Infrastructure as Code

    Terraform 3 environments, Docker Compose local, 15 CI/CD workflows.

Areas for improvement

  • Test coverage to strengthen

    Unit and E2E tests need to cover more critical paths.

  • Solo development

    One developer for an ambitious scope: real bus-factor risk.

What I would do differently

  • Incremental migrations from day 1

    Split the 98-model init migration, traceable and rollback-able.

  • Tests-first on regulatory modules

    Auth, billing, loan cases, audit trails: tests before UI.

  • Tighter MVP scope

    Broker + loan case + cifacil.io first, the rest unlocked by revenue.

  • A second engineer sooner

    Past the MVP, solo is a real risk on a regulated platform.

Lasting Lessons
  • 1.Regulated finance: map ORIAS, HCSF, GDPR, AML/KYC before the first line of code.
  • 2.Turborepo + pnpm monorepo: viable for multi-app SaaS thanks to shared packages.
  • 3.Centralized SSO: no duplicated auth logic, simpler sessions.
  • 4.IaC from day 1: 3 aligned environments through shared modules.

Related journey

Professional experience linked to this achievement

ACCENSEO - CTO & Founder

IT consulting and software company providing comprehensive services: custom software development, system integration, cloud infrastructure, database hosting, and technical advisory. Helping SMEs and mid-cap companies transform their IT systems into competitive advantages.

Skills applied

Technical and soft skills applied

Hard Skills

Architecture & Design5/5

System Architecture & Design

Designed monorepo architecture with 7 apps, 14 shared packages, and 98 database models

Software Development5/5

Full-Stack Development

Built 211K-line multi-app SaaS platform with Next.js 16, React 19, TypeScript 5.9, and Prisma 7

DevOps & Cloud5/5

Cloud Architecture

Terraform IaC with 3 environments, Docker Compose, Neon PostgreSQL, Upstash Redis, Cloudflare R2

DevOps & Cloud4/5

DevOps & CI/CD

15 GitHub Actions workflows, Docker multi-service, Traefik reverse proxy, automated releases

Image gallery

Project screenshots and visuals

Looking for a technical co-founder to bring your SaaS project to life?

You have a SaaS idea, a potential first customer, a strong domain intuition - but the technical half of the duo is missing. I co-founded and shipped this platform solo (7 Next.js apps, 14 shared packages, embedded AI, fintech-grade security, infrastructure as code) alongside a brokerage domain expert. If you are looking for a CTO co-founder able to own the architecture, the product and the production rollout of your SaaS, let's talk about your project.

Contact me