COST
< 0.10 USD per document
A pipeline of 3 chained AI agents that processes a complete document end to end.
Backend Developer · Focused on AI and Automation
I build backend systems and intelligent data pipelines using Python, FastAPI, and LLMs to transform unstructured documents into valuable business information.
01 / VERIFIABLE RESULTS
Every data point comes from a system I built.
COST
< 0.10 USD per document
A pipeline of 3 chained AI agents that processes a complete document end to end.
INPUT
3 formats PPTX · PDF · XLSX
A FastAPI REST API that ingests messy documents and turns them into structured data in PostgreSQL (JSONB).
DEPLOYMENT
In production
An AI invoice processing API with public Swagger, deployed on Azure with CI/CD.
02 / PORTFOLIO
AI pipelines and web systems, from idea to working demo.
PERSONAL PROJECT · IN PRODUCTION
Problem: PDF and text invoices are unstructured documents that are difficult to query and integrate with other systems.
Solution: A Python pipeline that extracts key data with LLMs (Groq), stores it transactionally in PostgreSQL, and exposes it through a REST API, with the flow orchestrated in n8n.
Result: API deployed on Azure with CI/CD and public Swagger. Open-source code.
PERSONAL PROJECT · AZURE DEMO
Problem: Much of the business intelligence is trapped in messy PPTX, PDF, and Excel files that cannot be queried.
Solution: A FastAPI REST API that ingests files, cleans them, and passes them through a chain of AI agents (Market Analyst → Brand Analyst). The result is stored in PostgreSQL (JSONB), and n8n generates insight dashboards.
Result: Processing a complete document with 3 chained agents costs less than 0.10 USD.
TEST DATA ACADEMIC PROJECT · UTN HAEDO · TEAM
Problem: Managing employees, branches, shifts, and attendance needs to be centralized in a web system.
Solution: A web system with an authenticated panel, manual and QR-code clock-in, location validation, an attendance dashboard, and PDF-exportable reports.
03 / APPROACH · DATA BEFORE AGENTS
Putting messy files into an AI chat and asking for a summary does not scale. Before connecting agents, the information must be cleaned and structured.
BEFORE · THE CHAOS
AFTER · THE FOUNDATION
Extract: A FastAPI REST API that ingests PPTX, PDF, and XLSX files.
Clean and analyze: deep text cleaning and chained AI agents (Market Analyst → Brand Analyst) with Groq.
Persist: structured storage in PostgreSQL with JSONB, ideal for auditing and fast searches.
Automate: n8n consumes the database and generates insight dashboards for consultants.
Garbage in, garbage out. Processing a complete document with 3 chained agents costs less than 0.10 USD.
View the project →04 / ABOUT ME
Principles I apply to every pipeline.
A REST API with FastAPI and Pydantic to validate what comes in and what goes out.
Transactional storage in PostgreSQL and JSONB so every result can be audited.
n8n orchestrates the flow from the file to the insights dashboard.
Deployment on Azure with GitHub Actions, not just code running locally.
UTN Haedo Regional Faculty
Average grade 8.46 / 10 · Graduate · Diploma in process View academic transcript05 / CONTACT
Tell me what you need. I can help you automate a process, organize messy data, or build an API that reaches production.