$ whoami
Marcelino Keyrouz
Software engineer & AI engineer
I build production systems end to end — from typed backends and infra to the model pipelines and agents that sit on top of them.
- epoch
- 0
- loss
- 4.820
$ cat about.md
I spend most of my time in the space between traditional software engineering and applied AI: designing the services, data pipelines, and infrastructure that make model-backed features reliable enough to run in production, then iterating on the models and prompts that sit on top of them. I care most about systems that are easy to reason about, test, and hand off.
$ git log --oneline --stat
- a3f9c1e2024 — Present
Senior AI Engineer
your-current-companyLeading design of retrieval and agent pipelines serving production traffic; owns evaluation harness and model-routing infra.
- 7d2b4aa2022 — 2024
Software Engineer
previous-companyBuilt and scaled backend services and internal tooling; introduced CI/CD and observability standards adopted team-wide.
- 1e8f0c52020 — 2022
Software Engineer
earlier-companyShipped customer-facing features across the stack, from database schema design to the React front end.
- 0a4d2f12019 — 2020
Software Engineering Intern
first-companyContributed to a data-processing pipeline handling millions of daily events.
$ ls ~/projects
Evaluation framework for retrieval-augmented pipelines — tracks faithfulness, retrieval precision, and regression across model versions.
PythonpytorchlangchainpytestReusable Terraform + CDK modules for standing up ML training and inference infrastructure on AWS.
TypeScriptterraformawscdkLightweight orchestration layer for routing tasks between specialized LLM agents with fallback and cost controls.
PythonfastapilanggraphredisStreaming metrics dashboard for model training runs — live loss curves, GPU utilization, and alerting.
TypeScriptqwikwebsocketsd3
$ cat skills.yaml