Principal Data Scientist
Sebastián
García Vilches
I build ML systems that move financial metrics. Currently leading data science at MACH, Chile's largest digital bank — from credit scoring to LLM-powered products. Bridging deep technical expertise with business strategy through an MBA at UC.
sebastiangarcia.dev
About
Building ML systems that matter
I'm a data scientist with 7+ years of experience building and deploying machine learning systems in fintech and banking. I specialize in the full ML lifecycle — from problem framing and feature engineering to production deployment and monitoring.
At MACH, Chile's largest digital bank, I lead technical direction for the data science team, translating business strategy into ML systems that reach millions of users.
I studied Civil Industrial Engineering at Pontificia Universidad Católica de Chile (top 10% of class), and I'm currently pursuing an MBA at UC to complement my technical depth with strategic leadership skills.
Experience
Where I've worked
7+ years building data products at the intersection of machine learning and financial services.
Technical lead for the data science team at Chile's largest digital bank. End-to-end ownership from research to production across recommenders, NLP, computer vision, and credit scoring systems.
- ▸Set technical direction for the data science org: modeling standards, feature engineering best practices, and production ML guidelines across multiple squads
- ▸Partnered with product, risk, and engineering leaders to prioritize high-impact ML initiatives, aligning technical roadmaps with business strategy
- ▸Built a personalization recommendation model increasing user engagement by +130% incremental clicks
- ▸Designed an LLM-based text classification model for the customer service chatbot, improving satisfaction score from 10% to 60%
- ▸Developed a Computer Vision identity validation model, replacing a costly external provider and reducing operational expenses by ~10%
- ▸Led credit scoring system development, enabling access to financial services for over 150,000 customers
- ▸Led a behavioral model enabling 80% credit exposure expansion for top-tier customers
- ▸Architected MACH's data lake migration to Apache Iceberg, reducing SQL query costs by 80% with full data versioning
- ▸Designed and institutionalized an org-wide MLOps monitoring framework for early detection of feature drift and model degradation
Impact
Impact by the numbers
Key results from production ML systems — measured, deployed, and monitored at scale.
Personalization recommendation model for mobile app home shortcuts.
LLM-based text classification transformed the customer service experience.
Data lake migration to Apache Iceberg with full versioning and scalability.
Behavioral model enabling significantly higher credit limits for top-tier clients.
Computer Vision identity validation replacing a costly third-party provider.
C-level financial report fully automated with near-zero calculation errors.
Skills
Tools of the trade
The technologies I use to build, deploy, and monitor ML systems at scale.
Languages
ML & AI
Data Engineering
Cloud (AWS)
MLOps
Projects
Side projects
Personal R&D — exploring the intersection of audio ML, LLM APIs, and developer tools.
Real-time Audio Transcription Pipeline
ActiveA personal productivity tool that captures voice input, filters silence, transcribes speech in real time, and passes the transcript to Claude for analysis — all with a push-to-talk interface.
Pipeline
- Real-time voice activity detection (VAD) using Silero to eliminate silence and reduce transcription costs
- Spanish-first transcription via Cohere's multilingual API
- LLM analysis layer (Claude) for summarization, data extraction, and compliance review
- Push-to-talk GUI designed as a voice-first interface for Claude Code
Contact
Let's connect
Open to new opportunities, collaborations, or just a good conversation about ML systems.
Reach me directly
I typically respond within 24-48 hours. For urgent matters, LinkedIn is the fastest way to reach me.