Converting a JavaScript Project with Flow Annotations to TypeScript
A practical migration plan for moving a JavaScript codebase from Flow annotations to TypeScript while preserving delivery speed, test confidence, and type-safety improvements.
Hello! I'm Octavian. I've been taking unfair advantage of CSS and JavaScript since 2006, and lately of other people's GPUs. As Director of Software Engineering at DOMSELARDI I ship systems, not slide decks. The two products below are live, and I run them.
Two live systems I design, ship and run at DOMSELARDI. Hosted in France, portable by design.
eIDAS · Tamper-evident
Prove what your AI did.
Evidentiary infrastructure that seals critical AI events into tamper-evident, eIDAS-timestamped proof bundles any auditor, regulator, or judge can verify independently.
Anti-spoofing · Multi-sensor
Spoofed drone signals, made visible.
Normalisation layer for the defence industry that turns raw satellite, ground-station and radio sensor streams into one signed, verifiable product - so drone signal spoofing becomes detectable instead of invisible.
Hello! I'm Octavian. I run software engineering at DOMSELARDI and I still write code. I'm lazy in the useful way: I learn new tools fast because I hate doing the same thing twice. The part of a project I like is the messy middle - vague requirements, brittle legacy, nobody sure who owns what, and systems that have to keep running while you change them.
I've built and reviewed production systems in ecommerce, adtech, banking, insurance, luxury, hospitality, medtech, public sector and space/sensor data. That's where I learned what breaks. It shows up now in how I design evidence layers, developer platforms and recovery plans.
I enjoy design patterns, modular reusable code and interfaces that hit 60fps. I hire and mentor engineers, speak at events and advise startups. What I care about: clear priorities, systems you can observe, and code that still makes sense to the next person.
I keep a study track running so these land properly, not superficially.
UiPath Academy RPA diplomas (2018); certification-track and professional-development coursework (Skillsoft), 2020-2021.
Picking a stack is the easy part. The useful part is finding the risks, writing down who guarantees what, and cutting a migration into steps that won't take production down. A team should be able to explain the platform during an incident or an audit.
How I work
Best-fit work
A practical migration plan for moving a JavaScript codebase from Flow annotations to TypeScript while preserving delivery speed, test confidence, and type-safety improvements.
Advice and hands-on delivery for teams working on AI systems, cloud platforms, security and legacy change.
I work with organizations that need to understand their own systems again. That means naming who owns what, finding the risk nobody has written down, and sequencing the migration.
On AI and data I focus on what actually fails in production: data ingestion, retrieval quality, model routing, evaluation and cost. Plus proof that a decision can be explained later.
On delivery: infrastructure as code, CI/CD, observability, monorepos and data access patterns, with engineering principles a team can actually follow.
Agent workflows, retrieval, model-serving pipelines, evaluations, and the cloud infrastructure automation that makes AI workloads repeatable, observable, and auditable.
Design and delivery of products where data location, export, audit trails, access controls, and verification are built in from the start.
I plan and deliver cloud migration programs with clear architecture decisions, migration sequencing, security controls, and cost discipline.
I assess risk, improve security, and help put practical controls in place for GDPR, incident response and work in regulated sectors.
Custom software, APIs, cloud-native platforms, observability, high availability, data layers, automation, and delivery tooling for sensitive workloads.
Migration and modernization of difficult legacy systems, turning executive goals into architecture, delivery governance, and changes teams can actually ship and adopt.
A delivery record across software engineering, cloud infrastructure, digital transformation, public-sector systems, and high-scale consumer platforms.
I set the technical direction at DOMSELARDI and I still write code, owning the architecture, the way we ship, security and the product direction. Recent work: an AI agent platform for a private equity firm taken from a single spec.md file to production; an eval harness that tests four production AI features against versioned golden datasets and stops a release if a score drops; hybrid retrieval on PostgreSQL (pgvector with an HNSW index fused with full-text search); AI governance that blocks approval while any generated number is unverified; the multi-cloud infrastructure behind the paid tier of a leading open-source agentic-development platform; and a defence sensor-data layer that makes drone signal spoofing detectable. I also write the rules for how AI coding agents work on our production code. I work in TypeScript/Node.js with Bun, React 19 on Next.js 16, Vue and SolidJS, Python with FastAPI and Celery, PostgreSQL/PostGIS, ClickHouse, MongoDB, Redis, pgvector, Kafka, Kubernetes with Helm and Terraform across AWS/Azure/GCP, AWS Bedrock, Playwright, and OpenTelemetry/Datadog observability.
I led distributed engineering teams of 10 to 30 across countries, migrating a bank's legacy finance suite and its 145,000 users to scalable cloud microservices. I contributed to 120+ technical hiring and senior-profile evaluations, improved delivery KPIs, grew the innovation budget by around EUR 4M, cut around EUR 2M of OPEX on an insurance product, ran the UiPath and Blue Prism RPA Centres of Excellence, and drove applied AI/ML work through Addstones. The stack covered Java/Django and Node.js/TypeScript services, Ruby on Rails, React/Angular/Vue/SolidJS frontends, PostgreSQL, Oracle, MongoDB, ClickHouse, Redis, Celery and Kafka pipelines, GraphQL, Kubernetes with Helm, Terraform and cloud-init across AWS/Azure/GCP, and Splunk/Datadog/Prometheus/Grafana/OpenTelemetry observability.
Led frontend architecture and team delivery for a B2B software platform, combining Angular, Python, testing automation, Elasticsearch, Solr, semantic technologies and RDF/OWL data models. Focused on maintainable architecture, reliable build systems, business workflows built on search, and practical engineering standards.
Built and maintained a complex B2B SaaS application using AngularJS, Ruby on Rails, Jenkins, Protractor, automated testing, and modern frontend tooling.
Built and integrated AngularJS infrastructure for a responsive cloud application used to manage multimedia storage for Orange customers. Delivered maintainable features, architectural improvements, and production fixes.
Delivered and integrated user interfaces for French brands and organizations, including Château Chenonceau, An Nissa, Coemy Group, Emeraude, Mademoiselle Fashion, Geomass, Foah Swim, SLA Paris, Devialet, Plateforme Fenêtre, and Cake l'Atelier.
Designed, coded, tested, documented, and supported business web interfaces for banking systems, including quality assurance and production support where reliability was critical. Over the same period, also delivered full-stack interactive-media and digital-publishing front-ends, including dynamic content builders and campaign systems.
Frontend engineering for high-traffic consumer and community products, plus senior frontend support for e-commerce interfaces and campaigns built to convert.
Built campaign websites, landing pages, newsletters, and custom advertising solutions reaching more than 8.2 million users each month. Collaborated on digital campaigns for brands including Vodafone, P&G, ING, HP, Adobe, Erste Group, and OTP Bank.
Created clear, conversion-oriented interfaces across desktop and mobile web. Worked with management and engineering through delivery.
Designed and developed web applications with JavaScript, HTML, CSS and PHP. Collaborated with designers on UI implementation and maintained ecommerce and presentation websites.
Excerpts from all 41 LinkedIn recommendations - senior leaders, clients, recruiters, and colleagues across software engineering, transformation, architecture, and complex technology delivery. Full texts on LinkedIn.
Cloud architecture, custom software, security, transformation or technical due diligence. Tell me what you are trying to do and what is in the way.