Cedric Vangout

Settleflow is a Stripe-like payment processor handling millions of payment requests per day across Europe, with a reconciliation layer running in the background ([settleflow.io](https://settleflow.io)). **Challenge:** Serve very high payment traffic under 15 ms latency, on a serverless stack, under PCI-DSS constraints. #### Approach Designed the event-driven architecture and cloud infrastructure for the payment processor, within a team of 5 engineers: - Event-driven architecture built on PostgreSQL CDC replication (aws_lambda extension) synced to AWS SNS and consumed by AWS Lambda functions - Webhook delivery to partners and billing/analytics metrics fed from the payment event stream - Infrastructure built entirely with Terraform modules and PCI-DSS-compliant AWS services (KMS, encrypted RDS, RDS Proxy, segmented VPC, Lambda, security groups, IAM) - Stripe-like public API with background reconciliation processing - VAMP Ratio (Visa Acquirer Monitoring Program) monitoring to track fraud and dispute ratios against Visa compliance thresholds - PSD2 Strong Customer Authentication (SCA) via 3D Secure for EU card payment flows, ensuring regulatory compliance - Moved non-critical SQL queries off the hot path into background jobs - Continuous PostgreSQL optimization: enabling and resolving slow query logs, monitoring CPU load (cache) and max connections (Lambda concurrency) - Built a CI/CD automation CLI (ARGC, Rust) deeply integrated with GitLab CI, deploying each MR to an ephemeral environment with a seeded database in under one minute - Git tag-triggered production deployments with automatic CHANGELOG updates (release notes extracted from commits) - Pipeline linter integration: OXC (TypeScript) and Squawk (detecting dangerous or slow PostgreSQL migrations in production) - Load regression tests (k6) run on every MR and release - Enforced clean code standards (SOLID, small reviewable PRs, OXC lint gate) across the team of 5 - Structured Lambda handlers with a hexagonal (ports & adapters) design, isolating payment domain logic from AWS-specific I/O for testability - Modeled the payment domain (transactions, reconciliation, webhooks) as bounded contexts to keep core business logic AWS-agnostic - Documented key architecture decisions (event-driven CDC pipeline, Lambda concurrency model) in ADRs #### Results - Served up to 100,000 payment requests per minute under 15 ms, with 900 concurrent Lambdas and an RDS Proxy pool fronting a scaled RDS instance - ~50x throughput improvement by choosing auto-scaling Lambdas over EC2 instances - Reduced SQL queries per application request from 150+ to under 15 through dedicated PostgreSQL indexes - Seeded ephemeral environments adopted by all contributors, accelerating iteration and regression review **Tech stack:** TypeScript, Bun, Hono, OpenAPI, Swagger Codegen, Rust, ARGC, OXC, Squawk, PostgreSQL (CDC, aws_lambda extension, RDS Proxy), AWS (Lambda, encrypted RDS, KMS, VPC, security groups, IAM, SNS), Terraform, AWS CDK, Bash, k6, GitLab CI/CD, PCI-DSS, PSD2, SCA, 3D Secure, OpenTelemetry

PrestaShop is the open-source e-commerce platform for creating and managing online stores; the project aimed to build its SaaS version ([onboarding.prestashop.com](https://onboarding.prestashop.com)). **Challenge:** Ship the SaaS version of PrestaShop after three failed attempts, coordinating four squads with conflicting priorities. #### Approach Led the "onboarding" squad (5 engineers) and built the SaaS version of PrestaShop, collaborating with the "core", "accounts", "billing", and "hosting" squads: - Designed and implemented a GraphQL gateway (graphql-yoga, graphql-codegen) federating 4 internal micro-services (onboarding, accounts, billing, hosting) under a unified BFF for the Vue frontend - Generated a type-safe GraphQL client from the schema, producing typed Vue hooks - GitHub Actions pipeline running GraphQL Inspector (graphql-inspector) to detect schema bugs, breaking changes, and regressions, plus tests, then publishing a Docker image to GCP Cloud Run - Proposed the initial architecture for onboarding.prestashop.com, jointly with the engineering manager - Wrote architecture decision records (ADR), enforced clean code standards via code review, and mentored junior developers in the squad - Led architecture meetings with tech leads from other squads to present the target architecture and arbitrate integration points and workflow dependencies - Aligned squad boundaries (onboarding, accounts, billing, hosting) with DDD bounded contexts to minimize cross-service coupling in the GraphQL gateway - Contributed to PrestaShop's open-source codebase to add generic extension points tailored to our needs - Reverse-engineered and contributed (MRs) to other squads' codebases to meet deadlines and shared objectives #### Results - Shipped ahead of schedule, onboarding 100+ merchants - GraphQL approach adopted by other squads in their codebases **Tech stack:** TypeScript, Node.js, GraphQL (graphql-yoga, graphql-codegen, graphql-inspector, federation, BFF), Vue.js, GCP (Cloud Run), Docker, GitHub Actions

elba is an all-in-one cybersecurity hub designed to secure teams in their daily work ([elba.security](https://elba.security)). #### Approach Embedded in a team of 5 engineers working directly with the CTO, responsible for developing a standalone feature, the "Training" module: - Delivered the "Training" module end-to-end: a 50+ step form assessing users' security knowledge (1,000+ users), consuming the Hasura GraphQL API and persisting progress incrementally, with React pages built on the internal design system - Migrated the application to an event-driven approach using Hasura event triggers - Built a Segment connector (Hasura event listener) forwarding relevant events — made seamless since every action was natively an event **Tech stack:** TypeScript, Node.js, React, PostgreSQL, Hasura, Segment, Vercel, GitHub Actions

Theodo FinTech builds innovative digital applications for the finance sector. #### Approach Embedded in a team of 3 engineers, contributing to internal administration tools (employee management portal): - Developed an employee profile management page within the internal portal - Extended the GraphQL API to cover new requirements - Set up an auto-generated GraphQL client via graphql-codegen - Added pages and components to the React codebase - Verified the Kafka integration to ensure proper event propagation for new features within the CQRS architecture **Tech stack:** Node.js, TypeScript, GraphQL, Apollo Server, React, MongoDB, Kafka, CQRS

PromTime is a data science company that compares patient health outcomes in real time ([promtime.org](https://promtime.org)). **Challenge:** Ingest high-volume patient health data following the complex PROM specification, under GDPR and HL7 constraints. #### Approach Embedded in a team of 3 engineers working directly with the CTO, selected for my GraphQL expertise to build a high-volume data ingestion API (PROMs): - Studied the PROM specification to propose a pragmatic GraphQL schema (300+ fields) bridging fixed and dynamic data - Built the GraphQL ingestion API (Apollo Server) with an interactive playground - Designed the GraphQL server in compliance with GDPR requirements and the HL7 standard - Wrote detailed documentation for partners: the PROM specification being complex, accurate field mapping is critical for business operations #### Results - 20 partners connected, collecting ~2,000 patient records/day - Security audit passed with no blocking findings **Tech stack:** TypeScript, Node.js, GraphQL, Apollo Server, PostgreSQL, GCP, Bitbucket, GDPR, HL7

jestocke.com is a peer-to-peer online marketplace for renting unused storage spaces ([jestocke.com](https://jestocke.com)). **Challenge:** Build the critical components of a growing marketplace with a two-person engineering team. #### Approach - Built the listing search engine (Algolia) and peer-to-peer booking and payment system (MangoPay) - Managed and monitored the production infrastructure (Heroku) - Helped structure the engineering team and recruited new members #### Results - 20K+ listings indexed, 100K+ transactions processed - Engineering team grown from 2 to 5 developers - Startup grown from 3 to 10+ employees through its second funding round (€2M) **Tech stack:** Python, Django, PostgreSQL, Heroku, MangoPay, Algolia