Field Guide · Ed. 01 · 11.2543°N 125.0000°E

Rolf Garces

Computer Science Student · Software Engineer · Researcher · Builder

Building technology that helps people navigate the real world through software, research, and entrepreneurship.

Explore the Journey View Projects
Rolf Garces
FIG. 00 · Expedition Lead

3+

Years as DOST-SEI Scholar

7

Projects Shipped

Top 6

APRU Int'l Hackathon · 70+ Teams

2

Leadership Roles Held

The Route · 10 Checkpoints

The Journey

Every skill I have was a checkpoint on the way here. The route runs from first principles to the systems I build today — and it keeps going.

  1. 01 · FOUNDATION · N 22° E · 2022–2023

    Programming Foundations

    Wrote first working programs; learned problem decomposition, debugging, and core language fluency.

    Code editor with early programming exercises on screen
    FIG. 01 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    Where the route begins: learning to think in code before thinking in frameworks. Fluency across languages, and the habit of decomposing a messy problem into something a machine can follow.

  2. 02 · THEORY · N 41° W · 2023

    Algorithms & Data Structures

    Implemented core data structures and algorithms from scratch; learned to reason in time/space complexity.

    Whiteboard covered in graph and algorithm sketches
    FIG. 02 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    The terrain gets technical. Understanding why one approach costs more than another, and how the right structure turns an intractable problem into a tractable one — the groundwork that later made graph-coloring research possible.

  3. 03 · DATA · N 33° E · 2023–2024

    Database Systems

    Designed and normalized relational schemas; wrote and tuned the SQL behind real project databases.

    Database schema diagrams in a notebook
    FIG. 03 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    Relational design, normalization, and query performance. Learning that most software problems are really data problems wearing a costume.

  4. 04 · DISCOVERY · N 58° W · 2024

    Research

    DOST-SEI Scholar; conducted graph-coloring research — hypothesis, experiment, honest measurement.

    Research materials and glassware on a lab bench
    FIG. 04 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    A DOST-SEI scholarship and graph-coloring research reframed building as a scientific act: form a hypothesis, test it, measure honestly, and let the result — not the ego — decide what ships.

  5. 05 · VELOCITY · N 27° E · 2024–2025

    Hackathons

    Top-5 regional finish, then Honorable Mention (top 6 of 70+ teams) at the international APRU Tech Policy Hackathon 2025.

    Team working on laptops during a hackathon
    FIG. 05 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    Compressing the whole arc — idea to demo — into a weekend. A Top-5 finish at the Regional Smart Communities Exposition, then an Honorable Mention at the APRU Tech Policy Hackathon 2025 in Google Bangkok — top 6 among 70+ teams from across Southeast Asia. Both taught more about scope, teamwork, and shipping than any amount of unhurried planning could.

  6. 06 · VENTURE · N 47° W · 2025

    Startup Building

    Validating startup concepts with real users — problem interviews and prototypes before code.

    Sticky notes and product planning on a whiteboard
    FIG. 06 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    The Startup Lab: treating ideas as hypotheses about people, validating with real users before writing much code, and learning that the hardest engineering is often deciding what not to build.

  7. 07 · CRAFT · N 38° E · 2025–2026

    Software Engineering

    Shipped production systems (Lab IMS, AIGRI) — maintainable code, reviews, team collaboration.

    Production code under review on a monitor
    FIG. 07 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    Beyond making it work — making it maintainable, reviewable, and collaborative. Real projects (IMS, AIGRI) where the code has to outlive the moment it was written.

  8. 08 · SCALE · N 52° W · 2026

    Cloud Computing

    Deploying to cloud infrastructure; learning the cost, reliability, and operations trade-offs of scale.

    Rows of servers in a data-center aisle
    FIG. 08 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    Infrastructure, deployment, and the economics of running software at scale. The current frontier — where an engineering decision becomes an operational one.

  9. 09 · FIELD WORK · N 44° E · 2026

    Internship — MYT SoftDev Solutions

    Production engineering at MYT SoftDev Solutions Inc. (Cebu City) — ERP performance work and a ticketing notification system for paying clients.

    Software development office workspace
    FIG. 09 · PLACEHOLDER — SWAP FOR A REAL PHOTO

    First professional deployment: inside MYT SoftDev Solutions' live codebases in Cebu City — profiling and optimizing two client ERP systems (~40% faster loads on the heaviest modules) and closing the notification loop in a support ticketing system. Full field reports: Expeditions 03 & 04.

  10. 10 · HORIZON · N 30° W · 2026 →

    Future Expeditions

    Open to engineering roles and collaboration — harder problems, new domains, real impact.

    What's next is uncharted on purpose. New domains, harder problems, and technology that helps people navigate the real world. Still exploring.

Field Reports · 07 Expeditions

Featured Expeditions

Projects written up as case studies, not repo listings — each one a problem explored, an approach chosen, and a lesson carried forward.

Expedition 01

Salin

AI-powered expense tracker for college students — 30+ registered users in its first week, fully organic.

  • JavaScript
  • HTML/CSS
  • PostgreSQL
Mission
Help college students actually keep tracking their money past week one, without the manual-entry friction that gets most budgeting apps abandoned.
Problem
Students already record spending somewhere — a notes app, a text to themselves — just never in the tracker itself, because re-entering everything by hand is the first habit to slip.
Approach
Built a mobile-first tracker that reads free-form notes and auto-fills transactions, so tracking happens where students were already writing things down instead of in a separate form. Added multi-card support and savings goals so the app models a real student’s finances, not one flat balance.
Outcome
30+ registered users within the first week of launch with zero paid promotion, and roughly 40% still logging expenses weekly a month in.
Lessons Learned
Removing one point of friction — manual entry — moved retention more than any feature added after it.

Expedition 02

Project AiGRI

Drone and computer-vision pipeline that turns crop damage into a submittable insurance claim — Honorable Mention (top 6 of 70+ teams), APRU Tech Policy Hackathon 2025, Google Bangkok.

  • Python
  • Flask
  • OpenCV
  • ResNet-50
  • React
  • PostgreSQL
Mission
Give smallholder farmers a faster, evidence-backed path to an insurance payout after crop damage, instead of a slow manual assessment.
Problem
Manual damage assessment is slow and inconsistent, and smallholder farmers often lack the documentation insurers require to process a claim at all.
Approach
Combined drone imagery with computer vision to run before/after damage analysis, packaging the results for LGU review and direct submission to insurance companies as a claim request.
Outcome
Honorable Mention at the APRU Tech Policy Hackathon 2025 (Google Bangkok) — top 6 among 70+ teams from economies across Southeast Asia — for a working pipeline connecting field imagery to a submittable, LGU-reviewed claim, cutting damage-assessment turnaround from weeks to days in pilot runs.
Lessons Learned
The hard part wasn’t the vision model — it was designing output that a non-technical institutional reviewer (the LGU) and an insurer could both act on directly.

Expedition 03

Production ERP Systems

Performance and feature work across two production ERP systems serving paying clients — software engineering internship.

  • PHP
  • CodeIgniter 3 & 4
  • MySQL
  • Hestia CP
Mission
Keep two clients’ ERP systems fast and reliable as their data volume and usage kept growing.
Problem
Data-fetching and page load were slowing down on live client systems as scale increased — a problem felt by real users every day, not a benchmark.
Approach
Profiled and optimized data-fetching and load performance, fixed existing bugs, and shipped client-requested features across two separate, actively-used ERP codebases.
Outcome
Roughly 40% faster page loads on the heaviest modules, on systems paying clients use daily.
Lessons Learned
Optimizing inside someone else’s production codebase is a different discipline than building from zero — every change has to respect existing architecture and can’t break what’s already shipped.

Expedition 04

Support Ticketing System

Ticketing upgrade with automated two-way notifications for clients and sales reps — same internship, second production system.

  • PHP
  • CodeIgniter 4
  • MySQL
  • Hestia CP
Mission
Make sure the right person is notified the moment a ticket needs their attention, instead of relying on someone remembering to check a dashboard.
Problem
Tickets logged without a notification loop just sat until someone manually checked — slow for clients and slow for the sales team responsible for the account.
Approach
Shipped email notifications to both the client and the relevant salesperson per project whenever a ticket was issued — in both directions — alongside data-load optimization and bug fixes.
Outcome
First response on client tickets moved from next-day to same-day once notifications closed the loop, for both clients and the internal sales team.
Lessons Learned
A notification system has two audiences with different needs (client vs. internal sales) — the same event had to be communicated differently to each.

Expedition 05

BiteSpot

Community-driven food discovery platform for Tacloban, staged on AWS for a real production deployment.

  • Laravel
  • PHP
  • MySQL
  • AWS EC2
  • AWS RDS
Mission
Give small, local Tacloban food spots — the kind that never show up on national delivery apps — a way to be found by their own community.
Problem
Local vendors have no shared discovery layer; word of mouth doesn’t scale past a neighborhood, and there’s no dedicated place for the community to surface and rate the spots they know.
Approach
Built a platform where vendors showcase their establishment and users rate, visit, and add new “bitespots,” then took it past the classroom by staging a simulated production deployment on AWS (EC2, RDS) to validate it as more than a school project.
Outcome
50+ local food spots cataloged during the pilot; started as a school project, now being evaluated for a real production launch.
Lessons Learned
Standing up real cloud infrastructure — even simulated — surfaced deployment and scaling questions a local dev environment never would have.

Expedition 06

Klaro

Mobile grade tracker that reads a syllabus PDF or screenshot and builds the semester’s grading breakdown automatically.

  • Flutter
  • Dart
  • SQLite
Mission
Keep students aware of exactly where they stand in a course while there’s still time to do something about it.
Problem
Grading breakdowns live buried in a syllabus PDF students read once and forget; nobody recomputes “what do I need on the final” by hand until it’s too late.
Approach
Built parsing that extracts grading weights directly from an uploaded syllabus — PDF or screenshot — turning a static document into a live per-semester tracker instead of something read once and closed.
Outcome
Adopted by 100+ students across the college within its first semester.
Lessons Learned
Real syllabi are inconsistent — different formats, different weighting language — so parsing had to be built for messiness, not a clean template.

Expedition 07

General Laboratory IMS

Desktop inventory system for UP Tacloban’s General Laboratory, replacing pen-and-paper borrow/return tracking.

  • Java
  • MySQL
  • JDBC
Mission
Remove the lost paperwork and guesswork from lab equipment borrowing, and hold both students and equipment accountable.
Problem
Paper-based tracking meant no reliable record of what left the lab, what condition it came back in, or who was responsible.
Approach
Built a desktop application where students log borrows and returns with condition proof, and admins get a searchable, real inventory instead of a drawer of forms.
Outcome
Adopted by the lab for day-to-day operations — tracking 400+ equipment items across multiple departments.
Lessons Learned
Designing for accountability meant the data model had to capture condition and proof, not just quantity — a different problem than a standard inventory CRUD app.

Field Notes · 07 Research Areas

Research & Innovation

Discovery is half the expedition. These are the areas under active study — each one feeding directly back into something built.

FIG. R-01

R-01

Active

Computer Vision

Damage quantification from drone imagery — classification pipelines built on pre-trained models (ResNet-50) and OpenCV, applied to real agricultural insurance claims rather than benchmark datasets.

  • EXP 02 · PROJECT AIGRI

FIG. R-02

R-02

Ongoing

Artificial Intelligence

How AI systems earn trust in institutional settings — designing model output that non-technical reviewers and decision-makers can act on directly, not just optimize for accuracy.

  • EXP 02 · PROJECT AIGRI
  • EXP 01 · SALIN

FIG. R-03

R-03

Ongoing

Machine Learning

Transfer learning and fine-tuning pre-trained models for domain-specific classification — making models work under small, messy, real-world datasets instead of curated ones.

  • EXP 02 · PROJECT AIGRI

FIG. R-04

R-04

Active

Graph Algorithms

Graph coloring — heuristic and exact approaches, and where the tractability boundary actually sits in practice. Hypothesis, experiment, honest measurement.

  • FIELD STUDY · GRAPH COLORING

FIG. R-05

R-05

Ongoing

Optimization

Performance under real constraints — algorithmic complexity treated as a budget, query tuning, and load-time profiling measured on live production systems.

  • EXP 03 · ERP SYSTEMS

FIG. R-06

R-06

Active

AI Applications

Free-form text into structured data — expense notes into transactions, syllabi into live grading models. Parsing built for how people actually write, not clean templates.

  • EXP 01 · SALIN
  • EXP 06 · KLARO

FIG. R-07

R-07

Ongoing

Software Engineering

What keeps production systems maintainable as they grow — architecture that survives other people's changes, and performance work measured on codebases that are already shipped.

  • EXP 03 · ERP SYSTEMS
  • EXP 04 · TICKETING SYSTEM

Instruments · 12 Ecosystems

Engineering Toolkit

The instruments carried on every expedition — clustered by ecosystem, not scored. What matters is what they've built.

SYS 01

×7

Languages

  • Python
  • Java
  • JavaScript
  • PHP
  • Dart
  • SQL
  • HTML/CSS

SYS 02

×5

Backend

  • Flask
  • Laravel
  • CodeIgniter 3 & 4
  • RESTful APIs
  • JDBC

SYS 03

×5

Frontend

  • React
  • Astro
  • Tailwind CSS
  • GSAP
  • Vanilla JS

SYS 04

×5

AI

  • TensorFlow
  • PyTorch
  • Pandas
  • NumPy
  • Transfer Learning

SYS 05

×4

Computer Vision

  • OpenCV
  • ResNet-50
  • Image Classification
  • Drone Imagery Analysis

SYS 06

×5

Databases

  • PostgreSQL
  • MySQL
  • SQLite
  • Schema Design & Normalization
  • Query Optimization

SYS 07

×5

Cloud

  • AWS EC2
  • AWS RDS
  • Docker
  • Hestia CP
  • Production Staging

SYS 08

×3

Mobile

  • Flutter
  • Dart
  • Offline-first Storage

SYS 09

×2

GIS

  • Drone Imagery Pipelines
  • Geospatial Data Handling

SYS 10

×5

Developer Tools

  • Git
  • GitHub
  • VS Code
  • IntelliJ
  • Docker

SYS 11

×7

Frameworks

  • Flask
  • Flutter
  • React
  • Laravel
  • CodeIgniter
  • TensorFlow
  • Astro

SYS 12

×6

Concepts

  • Data Structures & Algorithms
  • Object-Oriented Programming
  • Database Management
  • Software Engineering
  • Automata Theory
  • Artificial Intelligence

Dispatches

Writing

  • Engineering

    Optimizing Someone Else's Production Code

    Every change has to respect existing architecture and can't break what's shipped — a different discipline than building from zero.

    In draft

  • Engineering

    Parsing Real Documents: Building for Messiness, Not Templates

    What syllabus PDFs and free-form expense notes taught me about handling input the way people actually write it.

    In draft

  • Hackathon Lessons

    What a Weekend Hackathon Teaches That a Semester Can't

    Scope, teamwork, and shipping under a deadline — notes from a Top-5 regional finish.

    In draft

Live Log

Currently Exploring

  • Cloud architecture beyond a single EC2 box
  • PyTorch past the transfer-learning comfort zone
  • How startups validate before they build
  • Writing about engineering, in public

Off Duty

Beyond Code

  • Teaching

    Tutoring programming fundamentals at BrainBox — watching a concept click is still the best part.

  • Community

    DEVCON campus ambassador, organizing workshops that bridge students and industry.

  • Leadership

    Former Vice President, UP Interactive Society.

  • Off the map

    Exploring Tacloban's food scene — field research for BiteSpot, allegedly.

Next Expedition · Open

Ready to build something meaningful?

I'm looking for engineering roles, research collaborations, and people building things worth navigating toward. If that's you, the route is open.

Send a signal

Still exploring.

11.2543°N · 125.0000°E — Tacloban City