Selected work
Case Studies
Platforms I've architected and built — from AI-native SaaS and agentic systems to MCP infrastructure and enterprise-scale platforms.
AI-native SaaS · Current
An AI-powered SaaS platform that captures, structures, and analyzes organizational knowledge at scale. I own the architecture and engineering, and grew the team to 20+.
What I built
- Architected a horizontally scalable, multi-tenant platform across multiple clouds, cutting cloud costs through optimization.
- Built an agentic “Digital Expert” for contextual insights, backed by vector-based semantic retrieval and knowledge systems.
- Engineered a real-time AI interaction layer on LiveKit with event-driven services.
- Built high-throughput data pipelines (Python) for large-scale ingestion, transformation, and analysis.
- Integrated enterprise ecosystems — Slack, WhatsApp, Teams, Zoom — and led hiring and execution strategy.
- AI Agents
- RAG
- LiveKit
- Python
- Node.js
- pgvector
- AWS
- GCP
- Terraform
Real-time Commerce
A live-streaming commerce platform from the US where sellers run auctions and games for thousands of buyers. I architected the real-time platform and backend services, and led the engineering team.
What I built
- Architected and scaled a real-time, stream-based commerce platform supporting high concurrency.
- Built an in-house FAQ bot (Next.js, Node, LangChain, Pinecone, GPT) that improved after-sales service by 80%.
- Designed microservices (Node.js, Python) improving scalability by 70% and growing the user base by 50%.
- Built payments, logistics, and social-commerce services as API Gateway endpoints on Lambda, Kafka, RDS and DynamoDB.
- Led CI/CD transformation (Terraform, GitHub Actions), cutting deploy effort 3x, and mentored a 10+ team.
- Next.js
- Node
- Python
- GraphQL
- Kafka
- AWS
- Terraform
- LangChain
Industrial IoT · Real-time
An industrial IoT and fleet-monitoring platform. I built the real-time data systems and led the platform's move to multi-tenant microservices.
What I built
- Built real-time IoT and fleet-monitoring systems (Python, AWS IoT, Kinesis, Angular, Lambda) — helping attract a €10M investment.
- Designed event-driven backend systems (Python, Node.js) handling high-throughput device data.
- Migrated a monolith to microservices, improving reliability by 90%.
- Scaled the platform to a multi-tenant architecture supporting 200K+ users globally.
- Python
- Node.js
- AWS IoT
- Kinesis
- Lambda
- Angular
- Microservices
Knowledge Systems · MCP
PKS · PlanAct
A knowledge-intelligence platform that turns organizational data into interactive boards, decks, and answers — and exposes it all to AI clients over MCP.
What I built
- Built an MCP server (HTTP + stdio) exposing tools so AI clients can read and generate structured knowledge.
- Created React Flow + ELK auto-layout diagram boards and standalone slide decks as first-class content types.
- Implemented RAG vector QA over org files (pgvector) with SSE streaming via Postgres LISTEN/NOTIFY.
- Designed org multi-tenancy, per-user workspaces, and self-contained auth with platform/org admin.
- MCP
- React Flow
- ELK
- Postgres
- pgvector
- Next.js
- SSE
Enterprise Scale
Autodesk Forma (formerly Spacemaker) is an AI-driven urban-planning platform. I led backend and frontend architecture across the transition into Autodesk.
What I built
- Led backend transformation into serverless microservices (Python, Node.js) on AWS, reducing costs by 60% and scaling to 400K+ users.
- Architected data-intensive systems and APIs for AI-driven urban-planning workflows.
- Designed a micro-frontend architecture, improving development velocity by 60%.
- Led a component-library (Storybook) initiative, reducing frontend development time by 70%.
- React
- TypeScript
- Node
- Python
- Serverless
- AWS
- Micro-frontends
