Under the Hood

How the apps are built

• AppReact Native, Expo, TypeScript
• APIVercel serverless functions (Node.js)
• Data and sign-inSupabase: Postgres, phone sign-in
• RegisterSquare catalog, inventory, orders, payments

SheeshOvationShared

Sheesh Shawarma & CafeLive on the App Store and Google Play

• Payments verified server-side

  Card, Apple Pay and Sheesh Cash checkout. Before charging, the server re-prices every order from Square's catalog and re-checks store hours, the pickup slot, sold-out items, the owner's pause switch and the customer's live Sheesh Cash balance, so an outdated app can't order outside hours, at a stale price, or with cash it doesn't have.

• Loyalty that works in store too

  5% back on app orders, plus in-store credit by QR scan or phone number. Square webhooks attach the reward to the right account, and refunds reverse it.

• Two brands on one design-token system

  A side picker switches the whole app between Shawarma (dark) and Cafe (light), each with its own menu, hours and colours.

• Updates matched to each build

  Over-the-air updates are checked against each installed build's fingerprint, and owner controls show the exact running version.

Ovation Coffee & TeaComing soon to the App Store and Google Play

• Status earned from dollars spent, not points

  Redeeming never costs a member their tier. Pearl, Ruby and Noir earn at 1×, 1.4× and 2×, and status resets on each sign-up anniversary.

• App-only prices on the same register items

  Charged by the server, so the app can cost a little more without splitting the cafe's stock count.

• A crash traced into an open-source library

  Found through Sentry and traced to a scroll loop in a bottom-sheet library that overflowed the native stack. Shipped a guard while the upstream issue stays open.

• Live stock per location

  Read from Square for each of 6 locations: sold out at 1 or fewer, with quantity limits at checkout.

• Rewards and automatic nudges

  Tier-only rewards, birthday and welcome drinks, and push notifications for win-backs and status milestones.

Shared by both appsOne architecture, two clients

• Menu live from the register

  Menu, prices and sold-out items come from Square. The owner edits the register and the app updates with no release.

• Same-day releases

  EAS Build, Submit and Update ship fixes over the air without store review, with Sentry monitoring.

• Owner controls

  Owners pause ordering themselves, per location where there are several.

• Sandbox before production

  Payment and order logic is validated in Square's sandbox before real money moves.

Stack

AI, research and teaching

• M.S. Computer Science, AI/ML track

  Portland State University

  B.S. in Computer Science, summa cum laude.

• Fyxe: PII detection and anonymization

  Python, Microsoft Presidio, React

  Finds and removes personal information from prompts with NER and pattern detection instead of an LLM. Benchmarked against a RoBERTa baseline on ShareGPT52K.

• Clinical Informatics Research Intern

  Oregon Health and Science University (OCTRI) · 2023 to 2024

  Built a causal graph for disease quantification with propensity score matching and structural causal modeling, to analyze treatment outcomes in patient records.

• Software Research Intern

  Portland State University, Aphasia Lab · 2022 to 2023

  Built an acoustic analysis pipeline in Python with the Montreal Forced Aligner for post-stroke speech recordings, cutting manual analysis time by about half.

• COVID-19 predictive modeling

  Pandas, scikit-learn

  Random Forest regression predicting daily mortality rates (R² ≈ 0.74), with feature engineering and cross-validation.

• Graduate Teaching Assistant

  PSU Maseeh College of Engineering and Computer Science · 2024 to 2025

  Reviewed 100+ students' code through GitHub pull requests across three terms, and mentored in reinforcement learning, probabilistic models and heuristic search.