mchicao.dev
ProjectsMethodLabGitHub

Craft & carebeforeAI slop.

From gaming to Android modding, a lifelong curiosity now shapes how I build software: with judgment and attention to detail.

AI when it helps. Simple solutions when they suffice.

  1. Jun 11, 2012

    XDA · Android

    ROMs, kernels, and a Galaxy Young: learning by experimenting, modifying, and solving problems. Bricking my phone and deleting my IMEI taught me the essential value of backups (luckily, I had them).

  2. 2012-2023

    Skyrim & Minecraft modding

    Skyrim and Minecraft modding led me to start learning about programming. Venturing into Russian forums to learn how to play CS 1.6 on Steam for free taught me how to navigate unfamiliar software, troubleshoot, and solve technical hurdles on my own.

  3. Apr 20, 2014

    GitHub · bolmer

    I joined for LoL scripts. That curiosity led me to learn Lua, adapt code, and debug.

  4. Mar 01, 2019

    Curiosity for ML

    Breakthrough AI models mastering AlphaGo, StarCraft II, Dota 2, and AlphaFold, alongside channels like Two Minute Papers, sparked my passion for ML. I chose to study a broader engineering degree, opting to learn programming and ML on my own.

  5. Feb 24, 2025

    Dedicated Rig · LLMs

    Saved up to build a PC with dedicated VRAM to truly get hands-on with LLMs: running open-weight models locally, studying research papers, and actively working on experimental pre-training and post-training.

    Explore lab blog →

I've developed applications and tools for local use on Windows, Linux, macOS, Android/iOS, and web apps, using AWS and Azure.

WindowsWindowsLinuxLinuxmacOS / iOSmacOS / iOSAndroidAndroidAWSAWSAzureAzure

Your infrastructure. Your rules.

I know confidentiality is critical. My Open Sourced tools can run directly on your machines or secure network, guaranteeing your sensitive information is never shared with third parties.

code

Open source on GitHub

Some of my apps are open source: inspect the code, use them, and contribute directly on my GitHub.

open_in_newView GitHub

From bottleneck to operations in three steps.

01

Identify the bottleneck

Duplicate data? Reports that break on Mondays? Audits that take weeks? Each tool targets a specific problem.

02

Deploy the right tool

No month-long implementations. Activate the product that solves it, connect to existing infrastructure, and it starts working.

03

Scale operations

What starts solving a single problem becomes the team's operational backbone. Monitoring, traceability and automatic documentation.

Scalable and ready for deployment in enterprise production environments.

Projects designed to integrate with real infrastructure, operate with verifiable controls, and grow with the organization.

NiFi APIPythonPythonFastAPIFastAPINext.jsNext.jsRedshift

Data Governance Suite

Total visibility of data flows, risk auditing, and technical lineage

Comprehensive platform for Apache NiFi. Automates security backups, detects critical changes in SQL code, tracks field-level lineage, and sends impact alerts to Microsoft Teams.

  • Deep Lineage: Field-to-field data tracking
  • Smart Alerts: Microsoft Teams integration
  • Audit & Diff: SQL and Security (PII/Secrets) auditing
  • Impact Analysis: Blast Radius calculation
open_in_newView dashboard
Local LLMsRAGRedshiftFastAPIFastAPI

Data Governance Chat Bot

Instant queries about policies and metadata

Conversational assistant specialized in internal data structures, business rules, and governance policies. Find definitions without checking outdated wikis.

  • Interactive metadata chat
  • Lineage resolution in natural language
  • 24/7 support for analytics teams
PythonPythonAST ParsingNiFi APIAWSAWS SDK

Data Governance Auto Documentator

Automatic technical documentation generation

Autonomous value-stream documentor. Scans DBs, ETLs, Power BI DAX, Schedulers, AWS, and NiFi clusters to generate accurate data dictionaries and lineage maps.

  • Supports DB, ETL, PBI DAX, Schedulers, AWS, NIFI
  • Continuous technical glossary updates
  • End-to-end dependency mapping
RustRustGPUISQLiteSQLiteXLIFFTMX

LumenCAT

Assisted translation, local and under your control

An offline translation workspace for segment editing, translation memory reuse, and quality review without uploading documents to external services.

  • Segment editing and confirmation
  • Exact and fuzzy translation memory matches
  • Quality checks and automatic local saving
open_in_newView on GitHub
WhisperLocal AIWASAPI

Audio CC for meetings

Local live Closed Captions for meetings and translations

Real-time CC generation for any video calling platform. Facilitates comprehension and simultaneous translation with no visible latency.

  • Real-time CC generation
  • Platform agnostic (Teams, Zoom, Meet)
  • Total privacy via edge processing
RustRustDAG WorkflowsProvider CapsTelemetryegui
Open Source

SWARMS

Local, controlled orchestration for coding-agent workflows

A local-first runtime for choosing which models plan, code, review, and verify. Its deterministic Rust coordinator manages dependencies, provider concurrency, retries, resumable state, telemetry, and reports without locking the workflow to one vendor.

  • Configurable planning, execution, review, and verification roles
  • Up to 95.6% lower worker-agent cost in Cursor's experiment
  • Deterministic provider caps, retries, and workflow resume
  • Offline-safe defaults with opt-in real providers
  • Low-overhead native observability for active runs
open_in_newVisit SWARMSopen_in_newView on GitHubopen_in_newCursor research
PythonPythonDAXPower BILLMsllama.cpp

DAXBench

LLM evaluation through real Power BI execution

A benchmark I created in early 2026 to evaluate local and API models in one-shot and agentic modes, verifying their DAX measures through real Power BI execution.

  • 71 tasks with result verification in Power BI
  • Retries using real errors without revealing reference answers
  • Accuracy and latency comparisons with reproducible settings
open_in_newView dashboard
PythonPythonSource ASTSemantic IRPBIR/TMDLReactReact

DataVIZ · Verifiable BI Portability

Migrate dashboards across platforms while preserving intent, fidelity, and traceability

Evolves the Tableau → Power BI migrator into a platform that imports Tableau and Power BI assets, preserves their meaning in neutral intermediate contracts, measures fidelity gaps, and compiles the result to its own web runtime for SaaS or self-hosted deployment.

  • Tableau → Power BI migration with structural, numeric, and visual validation
  • Neutral IR for models, metrics, pages, visuals, and interactions
  • Fidelity reports classified as native, extended, or approximate
  • Web runtime and authoring SDK targeting SaaS and self-hosted deployments
open_in_newVisit siteopen_in_newView on GitHub
Next.jsNext.jsHealth APIsGamificationChartsAndroidiOS/AndroidWEB-Page

Mindsight

Physical and mental wellness dashboard focused on sleep

Mindsight

Personal health platform that measures, visualizes, and gamifies sleep and wellness habits. Turns health data into actionable metrics to improve quality of life day by day.

  • Sleep tracking and wellness metrics
  • Gamification with achievement system
  • Trend and pattern visualization
open_in_newVisit Mindsight
PythonPythonFlaskPostgreSQLPostgreSQLAzureAzure ADAWSAWS

PrivaSteward

Compliance platform for Chile's Data Protection Law 21.719

A platform I built for Transafety: it turns the obligations of Chilean law 21.719 into traceable, verifiable workflows: RAT documentation, organization and site management, data-subject portals, and audit-ready evidence.

  • RAT documentation and operation
  • Multi-organization with site-level isolation
  • Data-subject rights and rebuttal portals
  • Audit-ready evidence and traceability
open_in_newVisit PrivaSteward
PythonPythonFastAPIFastAPIspaCyReactReactSQLiteSQLCipher

DataDiscover

Automated discovery of personal data across corporate databases

A tool that runs inside the client's perimeter as a self-contained executable. It detects personal data via column names, content validators, statistical profiling, and NER (without moving the data) and drafts RAT proposals with an LLM for review in PrivaSteward.

  • Fully local multi-layer detection
  • On-premise executable, SQLCipher encrypted
  • LLM-assisted RAT proposals
  • Technical evidence integrated with PrivaSteward
PythonPythonFlaskPostgreSQLPostgreSQLAWSAWSGLM

SafeOps

Controlled diagnosis and remediation of infrastructure incidents

An engine I built for Transafety: it receives incidents from TranSafetycket (Dynatrace alerts), correlates them with AWS resources, and generates remediation plans with an LLM confined to a closed catalog. Every execution requires human sign-off and is persisted with its verification.

  • Incident correlation with AWS resources
  • Runbooks with mandatory human sign-off
  • Controlled, verifiable execution
  • HMAC-signed integration with TranSafetycket
ReactReact NativeExpoTypeScriptTypeScriptDynatraceDynatrace

OpsApp Android

NOC operations and observability on mobile

An Android app for Transafety that puts monitoring in your pocket: host health, service flows, incidents, AI chat, and root cause analysis over Dynatrace data, with SSO login and secure on-device storage.

  • Real-time NOC monitoring on mobile
  • AI chat and RCA over Dynatrace data
  • SSO login and secure storage

Engineering, not magic.

AI is a tool, not a religion. I use it where it genuinely reduces work and discard it where a simpler solution does the same job. What matters: the product works, is understandable, and is maintainable.

build

What I build

I build tools that solve real problems: data organization, smart search, automated auditing, measurable wellbeing.

block

What I don’t do

Sell you AI just to inflate the bill. Demos that only work in presentations. Adding complexity where a simpler tool would be faster, cheaper, and more reliable.

Contact me

matiaschicao@hotmail.com
mchicao.devData tools and applied intelligence, built by me.
Governance · Applied AI · Agents · Automation