Full-Stack Developer
Custom B2B web applications, end-to-end delivery, direct client interface, performance work and interactive experiences with Vue.js and Three.js.
- PHP
- MySQL
- Vue.js
- Three.js
> AI ENGINEER / LLM EXPERT
Stefano Facchin
01 / In numbers
Mywork,infournumbers.
- 8+ years of full-stack experience
- 4 companies and roles
- 6 areas of expertise
- 37 technologies in use
02 / Path
From code to agents.
- Adverteaser SRL 2018 - 2023
- HAVEIT SRLS 2019 - NOW
- Skuola Network 2023 - 2026
- JAKALA 2026 - NOW
Co-founder & IT Specialist
Technical ownership for Amazon Seller Central and Shopify e-commerce: scraping, pricing, supply chain and ML sales forecasting automation.
- Python
- Shopify
- Amazon FBA
- ML
- Scraping
Full-Stack Engineer & AI Expert
High-traffic education platforms and production AI systems with PHP/Symfony, React, Docker, Elasticsearch, LangGraph, MCP and GitLab CI/CD.
- PHP
- Symfony
- React
- Docker
- LangGraph
Senior Consultant - AI Expert Gen AI & NLP
Consulting on Generative AI and NLP solutions: AI agent orchestration, RAG systems and advanced agentic systems.
- Gen AI
- NLP
- RAG
- AI agents
- Agent orchestration
// interlude
Breaking news, every single year.
- 2012 “PHP is dead”
- 2015 “PHP is dead”
- 2018 “PHP is dead”
- 2021 “PHP is dead”
- 2024 “PHP is dead”
PHP, 2026: still in production.
It's still in my stack, next to Python and the agents.
03 / Method
How I take AI to production
Istartfromtheoperationalproblem,notfromtheloudesttechnology.
Idesignproduction-gradesolutionswithclearcontrolpoints:logging,tracing,metricsandfallbacks.
Imeasureperformance,scalability,maintenancecostandimpactbeforecallingworkdone.
// interlude
How almost every AI project starts.
- Client We need AI.
- Me Great. To do what?
- Client …AI.
- Me Ok. Let's start from the data.
04 / Today
Orchestrating agents, with the right data.
- 01 Documents
- 02 Chunking
- 03 Embedding
- 04 Retrieval
- 05 Agents
- 06 Answer
- 01 PDFs, pages and tickets enter the pipeline.
- 02 Texts are split into small, coherent chunks.
- 03 Each chunk becomes a vector that captures its meaning.
- 04 For every question, the closest chunks are retrieved.
- 05 Agents plan, call tools and check each other.
- 06 The answer ships with its sources, so it can be verified.
- Gen AI
- NLP
- RAG
- AI agents
- Agent orchestration
Best before
This diagram expires in
This diagram expires on 11 April 2027.
Shelf life left
The tools change every six months. The method does not.
// demo
Try it.
Two experiments, no LLM bothered: temperature changes the tone, retrieval finds the sources.
The same bio, at a different temperature.
I design AI agent orchestration, RAG systems and LLM automation, and use 8+ years of full-stack experience to take them from prototype to production.
At 0 a notary writes it, at 2 a bard.
Ask my RAG.
Real retrieval over the content of this page, answers written by me.
Pick a question
Retrieved chunks
Answer
- What stack do you use for RAG systems?
- For RAG I use Python with LangGraph, MCP and FastAPI, and Qdrant and Elasticsearch for search. Today I design them at JAKALA; before that I took them to production at Skuola Network.
- Where do you work now?
- Since September 2026 I have been Senior Consultant, AI Expert Gen AI & NLP at JAKALA: AI agent orchestration, RAG systems and advanced agentic systems.
- Do you build automation and scraping?
- Yes: n8n workflows and Python scripts for scraping, data sync and pricing. They grew out of the e-commerce business I have run with HAVEIT since 2019.
- Do you still work with PHP and Symfony?
- Yes. I started with PHP in 2018, and Symfony was my stack on high-traffic platforms at Skuola Network. It is still in my stack, next to Python.
05 / Projects
What I build.
-
AI & Search
AI Search / RAG System
RAG systems, semantic search and agentic workflows with LangGraph, MCP and FastAPI for advanced queries and less manual work.
- Python
- LangGraph
- MCP
- RAG
- FastAPI
-
Automation
Workflow Automation
Python automation and n8n workflows for scraping, data sync, pricing and repeatable operations across heterogeneous systems.
- n8n
- Python
- API
- Scraping
- Cron
-
Web
E-commerce Platform
Web applications and e-commerce platforms with PHP, Symfony/Vue, Shopify and MySQL, focused on performance, quality and delivery.
- PHP
- Symfony
- Vue.js
- Shopify
- MySQL
06 / Stack
37 technologies, 6 areas.
// interlude
Friday deploy, 5:59 PM.
$ git push origin main
▸ build ok · tests ok · lint ok
… almost there…
… just a sec…
… works on my machine
✓ online. Now let's talk.
07 / Contact
Let's talk
For consulting on AI agents and RAG systems, a professional opportunity, a product to build or a process to automate: tell me the context and the outcome you need.
contacts@stefanofacchin.comEnd credits
- Directed by
- Stefano Facchin
- Screenplay
- Stefano and an LLM, blaming each other for the bad jokes
- Cinematography
- 8 pixels per side
- Visual effects
- a CSS variable called --p
- Soundtrack
- the laptop fan
- Stunts
- the Friday deploy
- Deleted scenes
- a 3D city, lost to the Gommage
- Location scouting
- Zanarkand and Lumière
- Catering
- coffee, quantity ∞
- Special thanks
- PHP, for not dying
No servers were harmed in the making of this page.
// after the credits
Summary · Slot 01
Stefano Facchin
Lv. 8 (8 years of experience)
Senior Consultant - AI Expert Gen AI & NLP · JAKALA
- Chapters
- 8/8
- Interludes
- 3/3
- Projects
- 3
- Technologies
- 37
- Trophies
- 0/18
- Play time
- 00:00
Skills learned
- “Dead” technologies still pay salaries.
- Before “AI”, the question: to do what?
- Tools expire every six months. The method does not.
- Code is worth something when it holds up in production.