Vibecodeando tu Data Lake
AWS Community Day
An AWS Community Day talk on using AI-assisted (“vibe coding”) workflows to stand up a data lake, carrying the prototype-to-production mindset into cloud data infrastructure.
Hi, I am
Ramiro Savoie
I help teams ship
Information Systems Engineer from UTN-FRSF and postgraduate lecturer in the Master in Management & Analytics at ITBA. Twenty years of experience in Information Technology, ten of them applying Artificial Intelligence to large-scale data to turn it into informed business decisions.
I have worked across startups, corporations, universities, R&D and public administration, so I have solved problems of every kind. Behind many of those problems is data: for the last four years I have led deployr.ai, where we work alongside the data teams of Techint, MODO, Aeropuertos Argentina and Grupo Patagonia, among others, helping their data projects reach production.
About
Founder & Cloud Architect
Information Systems Engineer from UTN-FRSF and postgraduate lecturer in the Master in Management & Analytics at ITBA. Twenty years of experience in Information Technology, ten of them applying Artificial Intelligence to large-scale data to turn it into informed business decisions.
I have worked across startups, corporations, universities, R&D and public administration, so I have solved problems of every kind. Behind many of those problems is data: for the last four years I have led deployr.ai, where we work alongside the data teams of Techint, MODO, Aeropuertos Argentina and Grupo Patagonia, among others, helping their data projects reach production.
Experience
Founder & Cloud Architect
deployr.ai
A team of data engineers and scientists working as the technical partner of companies that need to turn their data into products. We help solve business problems with informed decisions, leaving installed capability with every client. We work alongside the data teams of Techint, MODO, Aeropuertos Argentina and Grupo Patagonia, among others.
Postgraduate Teacher
ITBA Innovation School
Teaching in the Master in Management & Analytics and in Executive / In-Company programs: LLMOps, Generative AI, MLOps, Text Mining and Time Series.
NLP Developer & Cloud Architect
Freelance
Independent contractor helping organizations extract value from their data through the design and implementation of data architectures, with a focus on MLOps.
Data Scientist
Xavia IoT
Design and implementation of an IoT platform. Worked on MediBox, a device for monitoring vital parameters of COVID-19 patients, and on precision-agriculture monitoring. Tools: Flask, InfluxDB, Grafana, AWS.
Specialist Teacher
Digital House
Led the ML Engineering modules of the Artificial Intelligence and Cloud Architect courses, teaching how to put models into production with AWS tools and containers.
Lead Data Scientist
Cliengo
Built the Machine Learning practice from zero. Data pipeline on AWS cleaning 20M+ conversations daily; NLP and topic modeling for intent discovery; classification models from LSTM to transformers (BERT/ALBERT).
Technical Leader
Garbarino
Redesigned the platform and processes of the Wedding Lists business unit on AWS; integration with branch systems, payment gateways, ERPs and mobile through backend services in SpringBoot and Node.js.
Early Engineering Roles
Multiple organizations
- Backend Developer at Avalancha (2015), TrackingTime (2014) and Bizonty (2013–2014).
- Fullstack Developer at the Ministry of Justice of Santa Fe (2011–2013).
- University Intern (2010) and Scholar (2009) at UTN-FRSF.
Education
Information Systems Engineer
UTN-FRSF
Research Scholarship (2009–2010), TICs Scholarship (2013) and FONSOFT thesis funding (2014).
Data Science
Digital House
Final project: guest-classification model for a hotel chain.
Deep Learning
Digital House
Neural-network architectures for text, image, audio and video. Final project: Intent-Based Bot.
Publications
Ramiro Savoie, Carolina Cinalli, “Atención al cliente 4.0: Los chatbots”, in Tratado de Inteligencia Artificial y Derecho, Thomson Reuters Argentina (2021). Book chapter on how chatbots work (intent detection, NLP and response generation) and their regulation, at the intersection of AI and Law.
Ramiro Savoie, “¿Qué viene después del model.predict()?” (poster), IA@Litoral 2023 — 2nd Litoral Artificial Intelligence Conference, sinc(i)–UNL–CONICET, Santa Fe (2023). MLOps case study: bringing a debt-recovery model into production for a Fintech (reproducibility, monitoring, testing and automation).
Writing
Agentes razonadores y uso de IA en las organizaciones: reflexiones desde la experiencia — ITBA Escuela de Innovación · EI NEWS #18. Reflections on reasoning agents and how organizations can use AI — strong for prototyping, analyzing user feedback and structuring product requirements; to be handled with care for high-precision financial or medical tasks.
Talks
A selection of talks, open classes and public conversations about applied AI, MLOps, cloud architecture, IoT and data products. The same throughline runs across all of them: turning technical ideas into systems that work in production and have real impact.
De Vibe Coding a producto escalable
Instituto Humai
With Maria Frances Gaska.
An open class about the step after the prototype: what happens when an idea validated with AI tools has to become a maintainable, scalable and operable product.
IA y Datos en Fintech: El camino de MODO hacia la hiperpersonalizacion
deployr
A conversation with Andres Scoccimarro, CDO of MODO, about data teams, engineering, MLOps, generative-AI use cases and hyperpersonalization in a high-scale fintech.
Como usar la Inteligencia Artificial para el desarrollo de Recursos Humanos
Camara Argentino China
With Andres Wei (Haomai).
A training on concrete AI applications in human resources: internal assistants, automation of frequent queries, recruiting processes and organizational adoption.
Clase Abierta: Conectar IoT con la Nube
Instituto Humai
With Luciano Savoie.
A practical introduction to connecting IoT devices with cloud services to turn physical signals into useful data for intelligent solutions.
Charla abierta: Introduccion a Machine Learning
Exactas UNSJ
An open lecture at the National University of San Juan (FCEFN): an introduction to Data Science, the kinds of problems Machine Learning is suited for, the challenges of taking on a project with these techniques, and a worked academic case. Available in two parts.
MLOps: Modelos de Machine Learning con impacto real en tu negocio
deployr
A talk on taking Machine Learning models out of the lab: lifecycle, operation, sensible scale, tooling components and practices connected to business metrics.
Electronica e IoT con MicroPython y Raspberry PI
Instituto Humai
With Luciano Savoie.
An open class on prototyping with electronics, MicroPython and Raspberry Pi, as a base for connected systems that capture data from the physical world.
¿Qué viene después del model.predict()?
IA@Litoral
Poster presented.
Poster and MLOps case study on taking a debt-recovery model into production for a fintech, focused on reproducibility, monitoring, testing and automation.
Lo bueno, lo malo y lo feo
HolaDataConf
A candid talk on what really happens when Machine Learning models reach production: the wins, the friction and the ugly parts of getting models to deliver.
Press
Los problemas que hoy acechan a las vacunas en América Latina — Infobae (2023). Named (“Ramiro Savoie, fundador de Deployr”) among the coordinators of a SLIPE/SLV think-tank on vaccine access, communication and trust in Latin America.
Lady IT · Encuentros Tech — ep. #11: Cloud Architect — Podcast (2022). A ~40-minute interview on cloud architecture, Machine Learning adoption, MLOps, data-career paths and the “human in the loop” concept.
Languages
Spanish (native) · English (professional)
Technical Focus
Tools and practices I use to move data and AI projects into production
Data & AI
Machine Learning
MLOps
LLMOps
Generative AI
NLP
Cloud & Backend
AWS
Docker
Python
Java
Node.js
Data Platforms
InfluxDB
Grafana
MongoDB
PostgreSQL
ELK
Contact
For consulting, teaching, talks, or collaborations around data, AI, cloud architecture, MLOps and LLMOps, use one of these public profiles: