Electrical engineering · telecommunications · USP

Engineering systems that can be measured, tested and trusted.

I work across AI/ML evaluation, signal processing, Android and cloud software. My projects emphasize reproducible experiments, privacy-aware design and careful validation on real systems.

São Paulo, Brazil · Portuguese (native) · English (fluent)

Signal, model and validation diagram A measured signal passes through an evaluation block and produces validated outputs. EVALUATE model · signal · system MEASURE → COMPARE → VERIFY reproducible evidence over intuition
6Android projects
1Telecom capstone
PT / ENWorking languages

Selected work

Mobile products built around hard constraints.

These projects span digital wellbeing, on-device speech AI, accessibility, document security, road safety and Android system integrations. The common thread is deliberate engineering: explicit privacy boundaries, device-level QA and measurable behavior.

Lost Time digital wellbeing illustration 3h 42m
01 · Digital wellbeing

Lost Time

An Android screen-time system that turns app usage into actionable limits, overlays, history and reports. Built in Kotlin with background services, local analytics, widgets and a Windows companion.

KotlinAndroidRoomWindows companion
Voice2Text waveform and transcription illustration
02 · On-device AI

Voice2Text

Private Android transcription for meetings and eligible media capture. Whisper runs locally, with local translation and optional on-device text processing through llama.cpp/Qwen models; audio and transcripts remain on the device.

Whisper.cppllama.cppKotlinLocal AI
Radar Brasil road-safety illustration 60
03 · Geospatial systems

Radar Brasil

An audio-first road-safety companion designed to coexist with navigation apps. Its data pipeline privileges official Brazilian traffic-authority sources, preserves provenance and rejects guessed coordinates, limits or directions.

AndroidGeospatial dataPython pipelinesOfficial-source QA
Better Notifications edge-lighting illustration
04 · Android rendering

Better Notifications

A configurable notification edge-lighting system with multiple rendering modes, app-matched color and privacy-first event handling. The core visual path reacts to notification events without consuming message content.

KotlinWindowManagerChoreographerPrivacy by design
Select Anywhere text-selection illustration COPY SELECTED TEXT
05 · Accessibility & OCR

Select Anywhere

A one-shot Android selector for text that other apps do not expose through their own selection UI. It combines Accessibility geometry with local OCR while protecting native gestures and keeping captured content on-device.

Jetpack ComposeAccessibilityLocal OCRUnicode geometry
Easy Sign document-signing illustration
06 · Documents & security

Easy Sign

A local-first PDF signing utility with vector ink, reusable signatures, structural PDF preservation and protected signing workflows. Its security track uses Android Keystore and region-gated ICP-Brasil support.

KotlinPDFBoxAndroid KeystoreVector graphics

Telecommunications capstone · USP

Real-time speech denoising & virtual microphone.

A collaborative engineering project that benchmarked classical and neural denoisers under a shared harness, then carried the selected signal path toward a Windows virtual-capture prototype.

  • Compared causal STFT, OM-LSA/IMCRA, RNNoise, WebRTC APM and DeepFilterNet3.
  • Used objective measurements and blind listening rather than a single quality metric.
  • Validated block timing and transport integrity, and documented where VM scheduling prevented a defensible end-to-end real-time claim.
Audio denoising signal chain illustration DENOISER benchmark · select VIRTUAL MIC transport · capture

About

Engineering with evidence as the interface.

I study Electrical Engineering with a Telecommunications emphasis at the University of São Paulo. My strongest work sits at the boundary between models and systems: designing experiments, interrogating failure modes, building the software around the model and turning results into clear technical decisions.

  • AI / model evaluation
  • Python & scientific computing
  • Signal processing
  • Kotlin / Android
  • Automated testing
  • AWS serverless systems