Harry De Torres

Software developer · Aparri, Cagayan, PH

--:-- GMT+8IT intern · since Nov 2026

I build software for what happens after the demo.1

I’m Harry De Torres, an IT student and intern from Aparri, Cagayan, in the Philippines. I build mobile products end to end, from the database on the phone to the rules on the server, and I spend most of my time on the cases a demo skips.

Currently building

As of 13 Nov 2026
Pre-launch · in active development

Upsenso

Getting the backend launch-ready: database tests that gate releases, a production database that can be rebuilt from the repository, and plan states that say what’s actually happening.

Bench logFrom the commit history
  1. Plan messages now say what’s happening: “plan still loading” instead of “not on your plan”, and a new device waiting for its first slot is no longer shown as a refusal.

  2. Production releases ship from main only.

  3. The production database can be rebuilt from the repository. Three storage buckets and eleven storage policies that had been created by hand are now migrations.

  4. Three database rules tightened, each with a regression test: product photos need product permission, logo writes are scoped to the business, and seat counts only count your own business.

  5. The pgTAP database tests now run in CI and gate releases. Sixteen files failed on the first run, so I quarantined them to get the gate live, then started repairing them and releasing them back in batches.

Selected work

3 case studies

Three products, built past the happy path.

Each has a full case study. The repositories are private, so the case studies show the parts of the code that matter, with file paths.

Independent product · Feb 2026 – present

Upsenso

A point of sale that keeps selling when the internet doesn’t.

Point of sale, inventory and back office for Philippine shops, cafés and restaurants. Every sale is written to an encrypted database on the device first and queued for sync, so the counter never waits on the network.

tables in an encrypted on-device database
83
commands per sync batch, halved when the server times out
25
test files, with 539 golden screenshots
777

Sole developer · Pre-launch · in active development

Read the case study
Branch sales report with a “Not yet synced” banner: one branch has sales or stock moves that haven’t synced, so its figures may be incomplete.
Sync queue: a product edit overwritten by a newer change and flagged for review, uploads that failed with their reasons, items waiting to upload, and a Retry failed button.
Sales overview: total sales, a daily bar chart, and breakdowns by payment method and by branch.
Real screens, rendered by Upsenso’s own golden tests · demo data

Team capstone · lead developer · Jul – Oct 2026

PalayCheck

A rice disease scanner that would rather say “not sure” than guess.

Farmers photograph a rice leaf and the phone checks the photo, checks that it is actually a rice leaf, and only then runs the disease model. Results the model isn’t sure about go to the Municipal Agriculture Office instead of showing a confident wrong answer.

Lead developer (app, backend, ML pipeline) · Not yet released · pilot scope: Aparri MAO

Read the case study
  1. Step 1

    Capture

    Camera or gallery photo, orientation fixed from EXIF.

    decoded in an isolate

  2. Gate 2

    Quality check

    Too dark, too bright or too blurry to judge?

    mean luma 45–215 · Laplacian variance ≥ 60

    If not: Retake, with the reason

  3. Gate 3

    Rice-leaf gate

    Is this a rice leaf at all?

    MobileNetV2 α0.5 · INT8 · 0.97 MB

    If not: Not a rice leaf

  4. Step 4

    Classify

    Crop to the leaf (+10%), then six classes.

    MobileNetV2 · INT8 · 2.7 MB

  5. Gate 5

    Confidence gate

    Sure enough to name a disease?

    top ≥ 0.45 · margin ≥ 0.10 · entropy ≤ 0.65

    If not: “Not sure”: report to the MAO

  6. Step 6

    Result

    Fuse several photos into one result.

    rejected photos never leave the phone

The scan pipeline, on the phone, with the thresholds the app ships with
Held-out test set
93.3%1,894 photos from the training sources, never trained on
Separate field dataset
75.7%382 photos, closer to what a phone in a rice field sees

Same model, same weights. On photos from a separate field dataset, accuracy falls to 75.7%. The case study shows where it fails, and why this version shipped anyway.

  • 1234

    Classic Strip4 photos

  • 1234

    Receipt4 photos

  • 123456789

    Photo Dump9 photos

  • 12345678

    Contact Sheet 358 photos

  • 1234567891011121314151617181920

    Twenty20 photos · 5 slides

Five of the 101 bundled templates, drawn from the app’s own template data · hover to fill the frames

Solo project · Oct – Nov 2026

Chroniva

A monochrome photo booth for Android phones and tablets.

It shoots a sequence, frames it in a template and exports a real composite strip. A carousel studio splits one wide design into Instagram slides that join back into a single picture.

bundled templates, 42 of them seamless carousels
101
renderer shared by the live preview and the export
1
test cases across 34 test files
364

Sole developer · In development · Google Play closed testing

Read the case study

Engineering journey

2023 – now

From school competitions to shipping products.

2026

  1. Nov 2026

    Milestone

    Started an IT internship

    In progress since 9 November.

  2. Oct – Nov 2026

    Project

    Built Chroniva

    First build from 31 October to 7 November, then an Android pipeline that ships to Google Play closed testing.

    Case study
  3. Sep 2026

    Project

    Rebuilt Upsenso’s interface as V2

    Moved every screen to the rebuilt interface in four phases, then deleted the legacy UI.

    Case study
  4. Jul – Oct 2026

    Project

    Led development of PalayCheck

    Team capstone: on-device rice disease scanning, and an outbreak map for the Municipal Agriculture Office.

    Case study
  5. May 2026

    Milestone

    Ledgify POS became Upsenso

    The point of sale I started in February got its current name.

  6. Feb 2026

    Project

    Started Upsenso

    An offline-first point of sale for Philippine shops, first called Ledgify POS.

    Case study

2025

  1. 29 Apr 2025

    Competition

    2nd place, General Programming

    ICT Competition at the 15th ICT Roadshow, Isabela State University – Echague.

    Certificate of Recognition for 2nd place in General Programming at the 15th ICT Roadshow 2025.

2023

  1. AY 2023–24

    Competition

    Web Development Champion

    With Louie Galamay as team Lovelace, College of Information and Computing Sciences.

    Poster announcing Louie Galamay and Harry De Torres, team Lovelace, as Web Development Champion.
  2. 2023

    Education

    Started BS Information Technology

    Cagayan State University – Aparri.

  3. 17 May 2023

    Competition

    2nd place, Web Design

    7th Inter-Highschool ICT Competition, hosted by Cagayan State University – Aparri.

    Certificate of Recognition for 2nd place in Web Design at the 7th Inter-Highschool ICT Competition, 17 May 2023.

Toolkit

By capability · no ratings

What I’ve used to ship the work above.

Grouped by what each tool was for. Instead of ratings, every row says where it was used, so you can check the work.

  • Mobile

    Flutter · Dart · Bloc / Cubit · get_it · go_router · React Native · Expo · TypeScript · Reanimated · Skia

    Used in: Upsenso, PalayCheck, Chroniva

  • Offline and data

    Drift / SQLite · SQLCipher · Outbox sync · Delta pull with watermarks · expo-sqlite

    Used in: Upsenso, PalayCheck, Chroniva

  • Backend

    Supabase · Postgres · Row-level security · RPCs · Edge Functions · Storage · pgTAP · Firebase Cloud Messaging

    Used in: Upsenso, PalayCheck, Chroniva

  • ML and AI

    Python · TensorFlow · TFLite INT8 export · Model evaluation · llama.cpp on-device · Cloud LLM behind an edge function

    Used in: PalayCheck, Upsenso

  • Web

    Next.js · React · TypeScript · Tailwind CSS · Vite · TanStack Query · Playwright

    Used in: Upsenso admin, PalayCheck site, This site

  • Delivery

    Git · GitHub Actions · Golden-test and pgTAP gates · EAS Build · Google Play testing tracks · Vercel · Claude Code · Codex

    Used in: Upsenso, Chroniva

  • Earlier work

    PHP · Laravel · C# · VB.NET · MySQL

    Used in: School projects

Journal

Beyond the build

Notes on the decisions behind the work.

Nothing is published yet. These are the drafts I’m writing, listed so you can see what’s coming.

  1. Draft

    Offline-first is a sync design problem

    Queuing requests is the easy part. Ordering, retries and replays are where an offline point of sale gets its books wrong.

    Upsenso · Sync · Postgres

  2. Draft

    Put a gate in front of the model

    A classifier always has an answer. Three checks that let a rice disease scanner say “retake”, “not a rice leaf” or “not sure” instead.

    PalayCheck · On-device ML · Flutter

  3. Draft

    93% on the test set, 76% in the field

    The same weights, two datasets, and the decision to ship a model that missed two of its own pre-registered criteria.

    PalayCheck · Evaluation · Machine learning

  4. Draft

    One renderer for preview and export

    How normalized template geometry and a single Skia composition keep a photo booth’s preview and its exported strip identical.

    Chroniva · React Native · Skia

Search the site

Home
WorkCase studies and smaller projects
JournalBeyond the Build, engineering notes
AboutStory, how I work, recognition
RésuméPrintable, one page
ContactEmail and links
UpsensoA point of sale that keeps selling when the internet doesn’t.
PalayCheckA rice disease scanner that would rather say “not sure” than guess.
ChronivaA monochrome photo booth for Android phones and tablets.
Copy email addressbox.devvv@gmail.com
Switch light or dark theme
GitHubgithub.com/DTHarryyy
LinkedInHarry De Torres
X@BoxDevvv
Book a 30-minute callCalendly