Rhenier Gustilo

Software Engineer: Machine Learning & Backend

Rhenier Gustilo

I build AI-powered software end to end: real-time speech pipelines on a Raspberry Pi, LLM services at the edge, and offline-first apps that keep working without a signal.

Education
BS Computer Engineering (Machine Learning specialization) · Polytechnic University of the Philippines, Sta. Mesa, Manila · September 2026 · President’s Lister
Availability
Based in the Philippines · open to on-site roles in the Philippines and remote roles
Core stack
  • Python
  • PyTorch
  • scikit-learn
  • LLM APIs
  • FastAPI
  • TypeScript
  • PostgreSQL
View projects

Selected work

Projects

Thesis · Team of 4 · 2025–26

EarPrompt

A wearable that listens to a conversation and whispers one short tip on how to respond.

The EarPrompt device: a white 3D-printed enclosure with a perforated lid and a microphone module on top, a front panel with one red and two blue buttons and a status LED, and a Bluetooth earbud case docked on the side.

Thesis prototype: a Raspberry Pi 5 earpiece that tells the wearer’s voice apart from the other speaker, generates a Taglish coaching tip with an LLM and streams it to Bluetooth earbuds, with a Flutter app for users and speech-language pathologists.

End-to-end latency
3.8 s avg 4.2 s P95 over 3 test sessions
LLM cost
$0.02–0.07 per 30-minute session

Role Lead developer — audio/AI pipeline, backend, device control and most of the mobile app

  • Python
  • PyTorch
  • LLM APIs
  • Raspberry Pi
  • Flutter

Live · Team of 4 · 2025–26

ConTeX

Photograph handwriting or drop in a PDF and get LaTeX back, compiled to a PDF preview in your browser.

Input and output of a ConTeX conversion. Input: a handwritten page reading "The energy of a particle is given by E = mc², where c is the speed of light in vacuum." Output: the LaTeX ConTeX returned, with the sentence as text and the equation as display math.

A Cloudflare Worker owns the Gemini prompt; if the AI is unavailable, OCR and an ONNX formula recognizer run in the browser instead. LaTeX compiles client-side with pdfTeX in WebAssembly. Ported from Flask to the edge with byte-identical output.

Automated checks
1,197 across 10 test suites
In-browser LaTeX compile
34 ms warm, desktop

Role Lead developer — edge port, security audit, test suites, CI/CD

  • JavaScript
  • Cloudflare Workers
  • Gemini API
  • WebAssembly
  • Firebase

Live · Solo · 2026

TindaBot

Tells sari-sari store owners what to buy, how many, and how much money to bring — offline, in Taglish.

Two TindaBot screens from the demo store. Left: the next supplier trip on Wednesday with about ₱4,450 to bring, ₱1,040 short of the owner’s budget, and a buy-now list. Right: the "Why?" sheet for Coca-Cola 1.5L, explaining the estimate from how often the owner buys it and saying it doesn’t know what’s left on the shelf.

An installable, local-first PWA: every entry is an append-only event on the phone, with optional Supabase sync under row-level security and an AI helper that only drafts entries for the owner to confirm. Business rules are tested against an independent Python reference model.

Automated tests
456 including 25 against a real Supabase project
Faster recompute
8× 383 → 46 ms on two years of data

Role Sole developer

  • TypeScript
  • React
  • PostgreSQL
  • IndexedDB
  • Gemini API

More projects

Live · Solo · 2026

TomatoSense

Classifies a tomato photo as ripe or unripe — one image or a batch of 20.

OpenCV preprocessing, PCA and an SVM (a linear kernel beat RBF and polynomial), served by FastAPI with single and batch endpoints and a plain-language accuracy page.

96.25% Accuracy — 80-image held-out test

  • Python
  • scikit-learn
  • OpenCV
  • FastAPI

Released · Solo · 2026

TextSnip

Drag a box over text in any Android app and get an editable copy, read on-device.

A floating bubble over every app. The selected region is captured with MediaProjection and read by Google ML Kit on the phone; the capture is deleted right after, and nothing is uploaded.

On-device OCR — no network needed

  • Flutter
  • Kotlin
  • ML Kit

Released · Solo · 2026

ClickMap

A Windows tray widget that turns keyboard keys into mouse clicks at saved screen points, system-wide.

A low-level keyboard hook on its own message-pump thread, SendInput/PostMessage click delivery, per-monitor DPI awareness, and atomic config writes with corrupt-file recovery.

  • C#
  • .NET 10
  • WPF
  • Win32

Internships

Experience

Before building software full-time, I deployed and supported real systems in hospitals, government buildings and offices.

  1. 2025

    IT & Security Systems Intern

    WJS Inc. (Webtek Advanced Technology and Security Solution)

    • Deployed and commissioned integrated security and building systems (CCTV, fire detection and alarm, public address, biometric access, LED walls and digital kiosks) for hospitals, government buildings and event venues.
    • Set up and troubleshot the network infrastructure behind them: LAN cable testing and termination, port addressing, and switch/router fault isolation.
    • Produced system design documentation (AutoCAD layouts, manuals and turnover reports) used for installation and client handover.

  2. 2024

    Technical Support Intern

    RVCom Computer Trading

    • Resolved hardware and software issues for 20+ clients a week using a systematic, step-by-step troubleshooting process.
    • Kept structured service logs to track recurring issues and improve turnaround.

  3. 2022

    Web Design Intern (Work Immersion)

    DOST – Philippine Nuclear Research Institute

    • Redesigned UI components of government web applications to improve accessibility.
    • Tested and debugged UI/UX issues before deployment.

Toolbox

Skills

Core stack

  • Python
  • PyTorch
  • scikit-learn
  • LLM APIs
  • FastAPI
  • TypeScript
  • PostgreSQL

Machine learning & AI

  • PyTorch (speech models)
  • scikit-learn
  • OpenCV
  • TensorFlow / Keras
  • NumPy · pandas
  • SpeechBrain speaker verification
  • Silero VAD
  • Praat / Parselmouth
  • ONNX Runtime Web
  • Tesseract · ML Kit OCR
  • LLM APIs (OpenAI, Gemini)
  • Prompt design & output guardrails
Used in

Backend & APIs

  • FastAPI
  • Flask
  • REST API design
  • WebSockets & streaming
  • Cloudflare Workers, Durable Objects, KV
  • Vercel serverless functions
  • Auth: Firebase Auth, Google OAuth, JWT/RS256
  • Rate limiting
Used in

Data & storage

  • PostgreSQL with Row Level Security (Supabase)
  • SQL
  • Firestore & security rules
  • IndexedDB (Dexie)
  • Event-sourced data models
Used in

Languages

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • Dart
  • C#
  • Kotlin
  • C++
  • R
  • Verilog HDL

Frontend & mobile

  • React · Vite · Zustand
  • Progressive web apps
  • Flutter
  • Android (MediaProjection, overlays)
  • WPF / .NET
  • WebAssembly integration
  • HTML/CSS
Used in

DevOps & quality

  • Git · GitHub Actions CI/CD
  • Vitest · Playwright
  • Python unit tests
  • Linux
  • Docker
  • Vercel · Render · Cloudflare · Firebase
  • Specs, architecture docs, validation logs
Used in

Embedded & hardware

  • Raspberry Pi 5 (GPIO, I²C sensors, audio I/O)
  • Arduino / ESP
  • Verilog HDL
Used in

Networking & systems

  • Switching, routing & wireless fundamentals
  • LAN cabling & testing
  • IP / port addressing
  • Cisco Packet Tracer
  • Security & building systems deployment

From internships and Cisco Networking Academy

Design & documentation

  • AutoCAD
  • Adobe Photoshop
  • Technical documentation

From internships

About & education

About

Portrait of Rhenier Gustilo

I’m a Computer Engineering graduate from the Polytechnic University of the Philippines, where I specialized in machine learning.

The part I enjoy most comes after the model works in a notebook: making it answer in real time on a Raspberry Pi, putting an LLM behind an API that can’t be misused, or keeping an app useful with no signal.

My internships in systems deployment and technical support showed me how things fail in the field, so my projects ship with tests, validation logs and fallbacks.

Education

BS Computer Engineering (Machine Learning specialization)

Polytechnic University of the Philippines, Sta. Mesa, Manila

Graduated September 2026 · President’s Lister

Thesis EarPrompt: A Real-Time Conversational AI Assistant for Enhancing Verbal Communication Team of 4 · advisers Engr. Rolito L. Mahaguay and Engr. Ferdinand O. Natividad

Certifications

Activities

  • Technical Committee, “Beyond the CpEctrum” seminar series, PUP Computer Engineering (2026)

Get in touch

Contact

I’m looking for software engineering, machine learning and backend roles, in the Philippines or remote. Email is the fastest way to reach me.

rheniergustilo06@gmail.com