# Georgios Hadjigeorgiou

_Amsterdam · Product & technology leader_

**Product thinking, engineering delivery, AI agents in production.**

I'm a product and technology leader, at home with strategy and roadmaps. I owned the roadmap for an API platform as a product manager, and now lead the engineers who build **Worldline's global APIs and merchant-facing apps**, and improve the developer experience for the clients who integrate with them. I also build agent tooling of my own. I run agents in production and evaluate them: tracing every run, tracking the details and running experiments to improve performance.

## Product and technology

### Product (Feb 2023 – Sep 2025)

Product Manager, API & Integrations

- Owned the roadmap for the [API and integration platform](https://docs.connect.worldline-solutions.com/), across 12 production applications.
- Launched Worldline's first public MCP server for merchants.
- Shipped API Explorer, multiple new payment products, SDKs, OpenAPI specs, MCP servers and real-time integration dashboards.
- Ran merchant calls to identify problems, assist with integration and settle product decisions.

### Technology (Sep 2025 – now)

Engineering Manager, API & Integrations

- Lead three squads of 24 engineers, owning the platform's architecture and reducing cloud costs.
- Leading the migration from AWS to GCP, and modernising the architecture.
- Built an in-house Slack agent that runs in production, and help my team with the AI adoption programme.
- Still work on the product side with design, on all roadmap initiatives.

I owned the roadmap as product manager, then moved into engineering delivery of the same platform.

## Experience

### Engineering Manager, Worldline

_Sep 2025 – now · Amsterdam_

Leading engineering for Worldline's global APIs, merchant-facing apps and the developer experience for clients.

Scope: 3 squads, 24 engineers, 5 hires so far.

- Runs agents in production and evaluates them: traces, run-level details, and experiments to improve performance.
- Works with product directors, the VP of Product, Security, Design, Data protection, Merchant Implementation and Sales.
- Sets platform standards and AI tooling adoption with colleagues.
- Partners with stakeholders to influence decisions on AI strategy.

### Product Manager, API & Integrations, Worldline

_Feb 2023 – Sep 2025 · 2 yrs 8 mos_

Product manager for Connect API: the REST APIs, SDKs and mobile-first interfaces merchants use to integrate. Presented Connect API on the Worldline Merchant Services channel.

### Co-founder, Product & Technology Manager, adoor

_Aug 2020 – Sep 2022 · Nicosia_

- Built adoor's iOS and Android apps with Flutter.
- Built and tested the API in Python.
- Integrated ordering, inventory and CRM with ERP software.

### Teaching Assistant, Researcher, TU Eindhoven

_Nov 2019 – Aug 2020 · Eindhoven_

Bachelor end project on the spatial and temporal consistency of brain networks across Multi-Echo fMRI combination methods. [Publication](https://www.scopus.com/pages/publications/85174819440#tab=citedBy)

## Education

### MSc Electrical Engineering, TU Delft

_2020 – 2023_

Signals and Systems track. TensorFlow, machine learning, signal processing.

### BSc Electrical Engineering, TU Eindhoven

_2016 – 2020_

Matlab, signal processing.

## Projects & talks

### MCP Server for AI agents (Featured · Worldline)

A translation layer between LLMs and platform APIs. Clients can use their agents to create payment links, do refunds, retrieve payment insights and manage payments in general.

- MCP
- REST APIs
- LLM integration

[Watch](https://youtu.be/SKKYEvaGzvA?si=xW-MR-BBM9iveW07)

### Connect API

REST tools, SDKs and mobile-first interfaces for simple, fast and secure integration with Worldline's platforms.

[Watch](https://youtu.be/Pa41wLgku-M?si=YUdW_1EbY5SeDxwU)

### Running and evaluating agents in production

Agents only improve when you can see what they did. I trace production runs with **Langfuse**, keep the details of each step, and run experiments comparing prompts, models and tools against the same cases, so changes are judged on evidence and performance keeps improving.

- Langfuse
- Tracing
- Evaluation
- Experiments

### Conversational agent for engineering operations

A Slack-native agent built in-house, in production. It **detects incidents**, runs scheduled tasks, runs root-cause investigation, informs the responsible tribe channels, and consolidates channel memory into shared knowledge every night.

- Python
- FastAPI
- Slack API
- LLM

### agent-sidecar

A detective, not a padlock, for AI coding agents. Six engines including a secrets vault, prompt-injection detector, policy engine and a Merkle-chained audit trail, so a bypass shows up as missing evidence. Self-audited before v0.1.0.

- Python
- FastAPI
- SQLite
- MCP

### Basalt

A macOS app that detects Teams meetings, records dual-track audio, transcribes on-device and writes a speaker-labelled Markdown transcript. No cloud transcription. Learns voiceprints from your corrections.

- Swift
- Core Audio taps
- Apple Speech
- Accessibility API

[Site](https://giorgos4.github.io/basalt-site/)

### adoor & academic research

Co-founded adoor: Flutter mobile apps on a Python API, integrated with ERP. Earlier, fMRI brain-network research at TU Eindhoven.

## Skills

### Domain knowledge

What I know about the problem spaces I build in. Payments is the deepest.

#### Strategy & roadmaps

- **Product roadmap**: Owned it for the API and integration platform as product manager, from 2023 to 2025.
- **Product strategy**: Defined strategic milestones to improve merchant developer experience and speed up integrations.
- **Technology strategy**: Set technology and architecture requirements for the platform's cloud, and drove buy-versus-build, microservices and database decisions.
- **Architecture standards**: Set platform, API and architecture standards that other tribes follow, and review other teams' platforms and recommend a direction.
- **Roadmap alignment**: Align priorities with Sales, Design and Customer Experience, and help customers through the launch of new products.
- **Business cases**: Defend ROI, budget and headcount requests to the people who control them.

#### Payments

- **Card payments & 3DS**: Improved the 3DS flow on hosted checkout, analysed exemptions to improve merchants' integrations, and integrated Visa AFT and multiple other payment product connections.
- **SEPA, Direct Debit & UK BACS**: Investigate disputed payments personally, tracing each one through validation logic and status timelines until the root cause is provable.
- **Global APMs**: Worked on Klarna, Troy, Samsung Pay, Hyundai Pay, Trustly and more.
- **Partial captures**: Led the API design for multiple partial captures without breaking changes.
- **Webhooks & payment events**: Drove the webhooks redesign on event streaming and autoscaling. Peak-season outages ended and there were zero incidents after launch.
- **Hosted checkout & drop-in UI**: Working with design on a new drop-in checkout method: competitor research, edge cases that break payment forms, simpler implementation.

#### Security & authentication

- **Passkeys / WebAuthn**: Defining requirements for passkeys as a checkout authentication method, an alternative to OTP and 3DS challenges.
- **Identity & access**: Standardised API authentication across the platform, on Keycloak and OpenID Connect.
- **Data governance**: Set how data is exposed through the natural-language query layer, together with security and data protection teams.
- **Vulnerability management**: Use Wiz and vulnerability tracking to make informed decisions on prioritising security improvements.

#### API & platform design

- **SDKs & OpenAPI**: Oversaw multi-language SDK generation and OpenAPI-led documentation.
- **Developer tooling**: Shipped API Explorer and worked on the Test Card Generator and Integration Dashboards, which cut merchant support escalations and integration friction.
- **Platform consolidation**: Assisted in merging payment platform offerings into one platform, migrated merchants off legacy systems and retired the overlapping ones.

#### Agentic commerce

- **Agent protocols (UCP, AP2)**: Contributed to Google's Universal Commerce Protocol and the AP2 and agentic commerce standards.
- **First European UCP payment handler**: Led the team that released it, analysed the protocol and held discussions with Google's tech team.
- **Visa & Mastercard**: Part of Visa's and Mastercard's initial agentic protocol discussions, and of discussions with clients to clarify needs and find product-market fit.
- **Public MCP server**: Launched Worldline's first public MCP server so merchants can build agent-based commerce workflows.

#### Commerce & product

- **E-commerce & ERP integration**: Co-founded a grocery delivery platform that connected a WooCommerce backend to Odoo, Softone, QuickBooks and several payment methods.
- **Funnel & conversion analytics**: Instrumented the funnel from landing to completed order, with deep links from campaigns into the native apps.
- **Customer discovery**: Ran hundreds of structured interviews with merchant payment PMs and developers. They shaped the webhooks redesign and the integration dashboards.
- **A/B testing**: Tested API Explorer layouts, payment product ranking on the selection page, and AI-written against human-written documentation.

#### Engineering leadership

- **Team leadership**: Lead three cross-functional squads of 24 engineers across frontend, backend and DevOps.
- **Hiring**: Hired engineers into the squads, owning screening, technical interviews and final decisions, with headcount matched to the roadmap.
- **On-call & incident response**: Improved the rotation, flattened the escalation policy, and adopted AI in the post-incident review process the squads use.
- **Stakeholder communication**: Present platform strategy, AI results and investment cases to senior leadership and merchant executives.
- **One-on-ones**: Bi-weekly, with coaching on ways of working: prioritising the day, managing time in meetings, and giving and receiving feedback.
- **Performance management**: Run a monthly performance cycle on GitLab and Jira delivery data, so feedback is continuous instead of annual.

### Technology

What I have used or shipped with, and what I did with it.

#### AI & agents

- **MCP servers**: Launched Worldline's first public MCP server as product manager, then built internal MCP integrations for ServiceNow, Jira, Tempo, Databricks and the AWS Console as engineering manager.
- **Claude Code & agent skills**: Assisting 24 engineers in the AI adoption journey. I lead by example: I build, test, deploy, monitor and improve agents in production.
- **Conversational agent (GC Tag)**: Built it myself in Python and FastAPI. It runs in production, detects incidents, investigates root cause, runs scheduled tasks, creates the Slack knowledge base and informs the owning channels.
- **Google ADK & Vertex AI**: Built the agent's sub-agents with the Agent Development Kit and deployed them on Vertex AI, with embedding and retrieval infrastructure.
- **Databricks Genie**: Wrote the semantic layer's table descriptions, SQL examples and join logic. Runs multiple queries a day for non-technical users.
- **RAG & vector search**: Built retrieval on Vertex AI and Databricks Vector Search.
- **Engineering automation**: Agents now do first-pass peer review, minor Spring Boot upgrades, security scans, incident investigation and time logging. About 8 engineer-hours a day saved on incidents, 4 real application vulnerabilities found.

#### AI quality & cost

- **Langfuse**: I run agents in production and use it to trace runs, track the details and run experiments that improve performance.
- **Evaluation**: Evaluation sets, accuracy scoring, LLM-as-judge and structured human review, with failures traced back to prompts, retrieval or the semantic layer.
- **Multi-model routing**: Route work across Claude, OpenAI and small open-weight models on cost and quality, and track token spend.
- **Open-weight models**: Fine-tuned Qwen with QLoRA on Apple Silicon (MLX) to generate Worldline SDK code, judged in phase-gated evaluations with a cost comparison against frontier models.
- **Agent security**: Built agent-sidecar solo: secrets vault, prompt-injection detector, policy engine and a Merkle-chained audit trail. I audited it myself and fixed the critical and high findings.

#### Languages & frameworks

- **Python & FastAPI**: GC Tag, agent-sidecar and data work.
- **Swift**: Built Basalt solo, a macOS app that transcribes meetings on-device.
- **Flutter**: Built iOS and Android apps at adoor.
- **SQL**: Authored the Genie semantic layer's reference SQL and validated its answers for correctness.

#### Cloud & infrastructure

- **AWS**: 12 applications on ECS, Lambda, Kafka, S3 and CloudFront. Squads run the instances.
- **Leading the AWS to GCP migration**: Leading the move, including CI/CD from Jenkins to GitLab. Earlier, helped move the platform from on-premises to AWS.
- **Terraform & Kubernetes**: Run and adjust Terraform configurations, deploy workloads and work on the CI pipelines that target containers.
- **DigitalOcean**: Provision and manage droplets for a personal automation project.
- **Kafka**: The event streaming behind the webhooks redesign.
- **Databases**: Drove PostgreSQL, Aurora and Oracle architecture investigations.

#### Delivery & observability

- **GitLab CI/CD & Jira**: Delivery data from both feeds the monthly team performance cycle.
- **Grafana, Kibana & Dynatrace**: Used for platform monitoring and incident work at Worldline.
- **CloudWatch**: Part of the AWS monitoring and alerting setup.
- **Integration dashboards**: Launched real-time dashboards for merchants' API and webhook logs and errors.
- **Google Analytics & Firebase**: Tracked acquisition, campaigns and the order funnel at adoor, with Crashlytics for app stability.
- **App Store Connect & Google Play**: Owned submission and releases for the adoor app, plus store listing optimisation and review responses.

### Languages spoken

- **Greek**: Native.
- **English**: Fluent.

## Contact

### Building with AI agents? Let's talk.

Based in Amsterdam. Open to engineers, AI enthusiasts and builders.

Email: [georgios@hadjigeorgiou.eu](mailto:georgios@hadjigeorgiou.eu)

- LinkedIn: <https://www.linkedin.com/in/georgios-hadjigeorgiou/>
- GitHub: <https://github.com/giorgos4>

### Interests

- AI agents & autonomous systems
- Agent observability & evals
- Model Context Protocol
- API design & DX
- Agentic commerce & payments
- Platform engineering
- AI adoption in engineering teams
- Signal processing & ML