Current role
Dec 2025 – Present
Current
PHINX Lab
· Remote
Team Leader & Senior Full Stack Developer
AI platform built on an orchestrator plus skills. Senior since joining; Team Leader since June 2026.
- I lead a team of 3 developers: technical planning, code reviews and architecture decisions, while staying hands-on in backend and applied AI.
- I'm the main author of two of the platform's services — a document analysis agent and an MCP server — where the hard problem isn't calling the model but making the system deterministic and operable.
- I designed a two-level orchestration architecture: a central orchestrator that dispatches to 7 structured-extraction skills, each with its own sub-orchestrator.
Three systems, three workflows
Folio01
Entre Ríos Education Council · 2023–2025
Family Allowance Certificate Upload and Validation System (Salario Familiar)
Province-wide: used by all school staff in Entre Ríos — teachers, principals, administrative and support staff. Upload, validation and tracking with full auditing and traceability, integrated with SAGE, the province's school management system. It includes a file-upload security layer written from scratch in plain PHP, with no external libraries due to a technical constraint set by the agency.
- Upload
- Validation 1
- Validation 2
- Payroll
- Approved
Rejected: back to upload
Resolved4 months → 5 weeks
Folio02
Entre Ríos Education Council · 2023–2025
Case File, Procedure and Internal Memo Management Platform
Covers virtually every case file the Council receives that is tied to payroll processes: thousands of case files, more than 10 departments and around 60 people. It isn't a case-file CRUD — it models the administrative workflow as a state machine, with its own authorization layer separate from SAGE's and managed from a dashboard, plus configurable departments, roles and workflow behavior for whatever changes with day-to-day operations.
- Upload
- Sent
- Received
- In progress
- On hold / Closed
Reopened: back to upload, keeping its history
Paperless10+ areas, ~60 people
Folio03
2026
This portfolio and its chatbot
The chatbot injects the full profile as context, with no RAG: the corpus fits in the context window, so retrieval would be over-engineering at this scale — more moving parts, more latency and a new failure mode in exchange for nothing. The decision is documented in the README. The LLM provider sits behind a swappable interface, with hard rules in the prompt, guardrail evals, rate limiting and a kill switch.
- Question
- Validation (Zod)
- Per-IP rate limit
- Prompt + full corpus
- Streaming answer
Provider failure: notice and retry
Publicauditable end to end
Reports from people I've worked with
“Either he reads minds, or he's a Lead who has the whole project and the next steps in his head. (…) The real MVP.”
Translated from Spanish.
“He stands out for his technical judgment and his ability to analyze problems and propose clear solutions. (…) I was able to see his growth and his profile for taking on greater responsibilities, especially in his move into a technical leadership role. I have no doubt Santi can add a lot of value to any team.”
Translated from Spanish.
“His analytical and problem-solving skills are extremely strong, and his knowledge is incredible: languages, frameworks, best practices, architecture and infrastructure, artificial intelligence. (…) Always willing to teach and to keep learning.”
Translated from Spanish.
“Besides being an excellent professional, he's a great person. He's always willing to help, share what he knows and go the extra mile when the situation calls for it. He's one of those people who make working as a team much easier and more manageable. Without a doubt, someone I would work with again and whom I fully recommend, both professionally and as a person.”
Translated from Spanish.
About me
The test of whether you understand something is being able to explain it simply enough for a kid to follow. If you can't get there, you haven't really learned it yet.