AI Engineer • Data Scientist

A decade in business. Now building the AI systems that run it.

After 10+ years growing businesses and leading analytics across 6 Latin American countries, I now build AI systems that respect data: local-first and open-source where privacy matters, Google Cloud where scale does.

View ProjectsView Certifications
Portrait of Jason Sepúlveda

Jason Sepúlveda

Master in Data Science

Pontificia Universidad Católica de Chile

Tech Stack

The tools I build with

Technologies I use to build modern AI systems, machine learning solutions, automation platforms, and data-driven applications.

AI & Agents

Google ADKOpenAIGeminiClaudeLangChainCrewAIMCPRAGVector DatabasesEmbeddingsSkillsN8N

Machine Learning & Deep Learning

Scikit-LearnPyTorchTensorFlowNeural NetworksNLPComputer VisionRegressionClassification

Data Analytics

PythonPandasNumPyMatplotlibPyPlotSeabornPySparkJupyterStatisticsData Visualization

Frontend

Next.jsReactTypeScriptTailwind CSSReflexStreamlit

Backend

FastAPIREST APIsPostgreSQLSQL ServerSupabase

Cloud & Data Platforms

Google CloudVertex AIDockerSAPPower BIETL PipelinesN8N
Featured Projects

Intelligent systems, end to end

Real, public repositories built around one thesis: local-first AI systems that run without depending on the cloud.

Local Speech AIOpen Source

whisper-loopback

Live, 100% local transcription of macOS system audio — no cloud, no API keys, nothing leaves the machine.

Problem

Transcribing live system audio (calls, meetings, media) usually means sending audio to a cloud API, trading privacy and latency for convenience.

Solution

Built a fully local pipeline: BlackHole captures system audio, whisper.cpp runs Metal-accelerated inference on Apple Silicon, and a LocalAgreement-2 streaming algorithm keeps transcription stable in real time — with automatic EN/ES language detection and Markdown/SRT/text output.

Pythonwhisper.cppfaster-whisperMetalBlackHoleTyper

Impact

Est. — for a meetings-heavy user (~5 hrs/week of calls): removes recurring cloud-transcription fees (~$250/yr at typical rates) and keeps 100% of audio on-device — a privacy/compliance win no cloud API can match. (123-test pipeline.)

GitHub
AI Intelligence MonitoringOpen Source

AI Observatory

A personal AI-ecosystem monitor that separates the developments that matter from daily noise.

Problem

Tracking meaningful progress across the AI ecosystem means wading through a constant, noisy stream of releases and announcements.

Solution

Built a daily collection and dedup pipeline with hybrid deterministic + local-LLM classification (Ollama, qwen2.5:7b) that produces traceable Markdown daily records, separating significant developments from noise.

PythonOllamaqwen2.5SQLitepytestuv

Impact

Est. — replaces ~30 min/day of manual AI-news triage with a traceable daily digest: ~2 hrs/week reclaimed (~90 hrs/year), at $0 in LLM API cost by running local models. (pytest-backed, source-traceable.)

GitHub
Knowledge Compilation EngineOpen Source

OpenKOS

A local-first knowledge compilation engine that turns raw sources into a compounding, queryable knowledge base.

Problem

Knowledge work usually means re-reading the same raw sources over and over instead of compounding what has already been learned.

Solution

Implements Karpathy's LLM Wiki pattern using the Open Knowledge Format (Markdown + YAML) — ingest, query, and lint commands built around provenance, citations, and freshness linting, with immutable raw sources and rebuildable indexes.

PythonOllamaMarkdownYAMLuv

Impact

Est. — turns repeated source re-reading into a queryable base built once and reused: ~2–3 hrs/week saved re-finding known material, with provenance and citations intact and $0 per-query cost (runs local). (12+ docs incl. 6 ADRs.)

GitHub
Agentic AIIn Development

Agentic System (Working Title)

An in-development agentic AI system extending the local-first thesis from transcription and knowledge tooling into autonomous workflows.

Problem

The local-first, open-source systems built so far handle transcription and knowledge — the next gap is autonomous, multi-step workflows that use them.

Solution

Designing an agent architecture on Google ADK to orchestrate planning, tool use, and memory across the existing local-first tools.

Google ADKGeminiOpenAIPython

Impact

Target — orchestrate the existing local-first tools into autonomous, multi-step workflows, compounding the per-tool savings (transcription, monitoring, knowledge) into hands-off pipelines. Today: architecture and tool selection underway.

How I Work

From business problem to measured result

Analyze & Consult

I start with the business, not the technology. Drawing on 10+ years operating across Latin America, I diagnose where the real impact is, quantify the opportunity, and define what success looks like — before a line of code is written.

Design the Solution

I turn that diagnosis into a solution designed for the outcome: the right architecture for the problem — local-first where privacy matters, Google Cloud where scale does — so the system serves the goal, not the hype.

Build & Measure

I implement the system end to end and measure it against the goals set up front — so impact is proven with numbers, not just claimed.

About

Bridging business strategy and technical execution

Portrait of Jason Sepúlveda

Jason Sepúlveda

Master in Data Science

Professional Evolution

  1. Business
  2. Digital Transformation
  3. Data Analytics
  4. Data Engineering
  5. AI Engineer & Data Scientist

I am a business-oriented AI Engineer and Data Scientist with more than 10 years of experience across analytics, customer experience, digital transformation, operations, and marketing throughout Latin America.

My career has allowed me to work at the intersection of business strategy and technology, helping organizations identify opportunities, solve complex problems, and generate measurable impact through data and AI.

Today, my focus is on building intelligent systems powered by machine learning, generative AI, agentic architectures, and automation platforms that improve how businesses operate and make decisions.

AI EngineeringData ScienceBusiness StrategyMachine LearningGenAIAutomation
Experience

A decade across business, data, and AI

A professional journey combining business leadership, analytics, digital transformation, and artificial intelligence.

  1. 1

    Business & Growth

    Partnerships Development Manager · Uber

    Sales Director · Comunidad Feliz

    Business Growth · Customer Acquisition · Partnership Development · B2B Sales · Revenue Growth

    • 1-point growth in market share
    • Led a 5-person partnerships team
    • Delivered on annual sales targets
    • New key B2B accounts secured
  2. 2

    Digital Transformation

    Software Field Specialist · Hilti Chile

    Digital Transformation · Process Optimization · Technology Adoption · Operational Efficiency

    • Process optimization initiatives
    • Workflow automation opportunities
    • Operational efficiency improvements
    • Technology adoption acceleration
  3. 3

    Analytics & Decision Intelligence

    Regional eCommerce & Analytics Manager · Hilti LATAM

    Analytics · Digital Products · Data-Driven Growth · Decision Intelligence

    • 35% growth in eCommerce sales
    • 80% increase in user adoption
    • LATAM became the top-performing region globally
    SQLPower BISAPGoogle AnalyticsData Visualization
  4. 4

    Data Engineering & Automation

    Regional CX Order-to-Payment Project Manager & Data Analyst · Hilti LATAM

    Data Engineering · Automation · Customer Experience · Business Intelligence

    • ETL pipelines built using Python
    • SAP data integration
    • Analytics delivery across more than 300 sales territories
    • 20% improvement in customer experience KPIs
    PythonSQLSAPPower BIETL Pipelines
  5. 5

    AI Engineer & Data Scientist

    Present

    Machine Learning · Generative AI · Agentic Systems · AI Applications · Intelligent Automation

Certifications

Always learning, by design

Continuous learning has been a core part of my professional growth, with more than 60 certifications completed across artificial intelligence, machine learning, cloud technologies, software development, and data science.

60+

Certifications

10+

Learning Platforms

10+

Years of Continuous Learning

Featured Certifications

AI Engineering

Scrimba

Machine Learning

DeepLearning.AI

Advanced Learning Algorithms

DeepLearning.AI

Claude Code 101

Anthropic

Introduction to Model Context Protocol

Anthropic

Design, Develop and Deploy Multi-Agent Systems with CrewAI

DeepLearning.AI

Retrieval Augmented Generation (RAG)

DeepLearning.AI

Foundation: Introduction to LangChain

LangChain

View All Certifications

Learning Domains

AI & Agents

Building intelligent systems powered by LLMs, agentic architectures, orchestration frameworks, and retrieval systems.

AI EngineeringClaude CodeMCPCrewAILangChainRAGAI AgentsOpenAI APIsEmbeddingsVector DatabasesHeyGen

Machine Learning & Deep Learning

Developing predictive systems using statistical learning and deep learning techniques.

Machine LearningAdvanced Learning AlgorithmsSupervised Machine LearningUnsupervised LearningDeep LearningCNNsNLPLogistic RegressionDecision TreesRandom ForestComputer Vision

Analytics & Data Science

Transforming data into insights and decision intelligence.

Foundations of Data ScienceData Analysis with SQLIntroduction to Data AnalyticsPython for Data ScienceSQL DatabasesData Science ProjectsWeb ScrapingAnaconda & Jupyter

Software Engineering & Development

Building scalable applications and APIs that power modern AI systems.

FastAPIDjangoDjango REST FrameworkUnit TestingGit & GitHubSupabaseSupabase Advanced

Cloud & Infrastructure

Deploying and operating modern AI and data systems.

Google Cloud Digital LeaderDigital Transformation with Google CloudInfrastructure Modernization with Google CloudAI & Machine Learning on Google CloudDocker
Contact

Interested in AI, machine learning, automation, or data-driven transformation?

Let's connect and explore how intelligent systems can create business impact.

Or connect with me on