Multi-agent orchestration
Several agents, each with one responsibility, coordinated with Pydantic AI and wired into the systems and infrastructure the company already runs.
Agentic AI Developer · Madrid, Spain
Daniel
Hernández Gil
I build systems where several agents each own one job — and I keep them running. Ask the orchestrator below.
orchestrator.run
resume_agent
└ fetch_resume
work_agent
└ list_roles
└ rank_impact
media_agent
└ pipeline.stats
synthesize
1.84s total4 agents4 tools1 llm call
orchestrator output
9 yrsshipping software
2 yrsagents in production
4modalities orchestrated
2024 — present
What I’m building
At Game Strategies I’m building a new product around a set of agents with separate responsibilities, coordinated with Pydantic AI and connected to the infrastructure the company already runs.
Generative media pipelines
Text, image, video and 3D models combined into a single workflow, plus the backend and frontend around it.
Shipped, not demoed
Claude Code and Codex daily, from first prototype to the code that ships, against delivery dates agreed with the business.
Selected work
Things I shipped
These run inside a company, so the repositories stay closed. What follows is the shape of each system and what I owned in it — happy to walk through the architecture in a call.
Trellis2 APIprivate
Inference API that turns a single image into a 3D model, running TRELLIS.2 on H100 GPUs. Synchronous calls for small jobs, a queue for everything else.
Calypsoprivate
Android app for streaming live sports: camera control from the phone and real-time overlays drawn on top of the video feed.
Web Motion Atlasprivate
Catalogue of web motion and layout patterns that composes a chosen pattern into a precise prompt for a coding agent.
2017 — present
Nine years, on one axis
Each bar is a role placed on the same timeline. Read it top to bottom and you get the short version: test automation, then the platform underneath it, then the agents on top.
Agentic AI Developer
Game Strategies2024 — present- Building a new product around several agents with different responsibilities, coordinated with Pydantic AI.
- Built the pipelines that combine text, image, video and 3D models into one workflow, and the backend and frontend around them.
- Coding agents daily to take ideas from concept to a working MVP, on the dates agreed with the business.
SRE & Tooling Engineer
Game Strategies2022 — 2024- Internal tools that gave developers their own view of Kubernetes — logs, service state, deployments — so they stopped escalating to diagnose.
- Automated the whole product's deployments with Argo Workflows, Argo Events, Docker and Ansible, replacing steps still done by hand.
- Ran the infrastructure across cloud environments: networking, monitoring, production troubleshooting.
QA Automation Engineer
Game Strategies2020 — 2022- End-to-end suites for a microservices architecture in Python and JavaScript, taking coverage up by 80%.
- Automated API checks in the CI pipelines with GitHub Actions, and fixed the defects directly in backend and frontend code.
Full Stack Developer
CEITA SL2018 — 2020- Java, BPM and PostgreSQL for a banking client.
- Built a test automation framework with Java and FitNesse, and handled releases across frontend, backend and database.
Software Engineer
Indra2017 — 2018- Frontend and backend with Java and REST services, in a shared codebase with a large team.
Core stack
What I reach for
AI & agents
Build
Platform
Quality
Got an agent
that needs to ship?
I’m most useful where agents have to touch real systems and stay up afterwards. Send me the problem and I’ll tell you straight whether agents are the right tool for it.