Hwajong (HJ) Jeong
I craft software that
works.
Software Engineer — Mobile/Web & ML/AI
Shipping production software across the stack — from native mobile and web platforms to machine learning pipelines.
About
Engineering team lead at The Home Depot, where I drive the development of consumer-facing mobile applications used by millions. I lead multiple cross-functional teams shipping native Android and iOS features at scale — from architecture decisions down to production rollouts. Outside of work, I build across the full stack and dig into ML/AI research through Georgia Tech's OMSCS program.
Georgia Institute of Technology — M.S. Computer Science, AI Specialization (Expected May 2026)
Projects
The Home Depot Mobile App
Lead engineer for two cross-functional teams — one with 16 members, another with 11 — within the org that owns Home Depot's mobile app on the App Store and Google Play. The app serves 1.6 million daily active users across iOS and Android.
Home Depot Local Mobile AI
An on-device iOS AI feature built during The Home Depot's Innovation Week that takes a DIY customer's plain-language project description and returns a structured triage assessment: project category, difficulty rating, DIY feasibility, and routing across three revenue paths (materials purchase, tool rental, or Home Services booking). The full pipeline runs entirely on-device with zero cloud dependency, using a Qwen3-4B model fine-tuned with QLoRA on 5,000 synthetic examples, RAG over home improvement guides, and GBNF grammar-constrained inference to guarantee valid JSON output. Demonstrated live on a physical iPhone in airplane mode.
Home Depot Watch App
An in-house Apple Watch companion app that brings core Home Depot functionality to the wrist — quick order status checks, store aisle lookup, shopping list access, and barcode-ready product details without pulling out a phone. Built with SwiftUI and WatchKit, syncing state with the main iOS app via Watch Connectivity. Currently in closed beta with internal teams and select store associates.
Home Depot Instore Navigation
An AR wayfinding prototype built during Home Depot's Innovation Week that overlays turn-by-turn navigation directly onto the store floor through the camera feed. Uses ARKit and PointrKit for indoor positioning and spatial tracking, rendering real-time directional arrows and distance markers anchored to physical aisles. Designed to embed directly into the existing Store Mode experience so customers can tap a product in their list and get guided to the exact bay — no map reading required.
GlamBot — GATech Research AI
A locally-hosted RAG chatbot built for Georgia Tech's Educational Data Mining Research Lab, helping new researchers onboard to a 90.4M-event edX clickstream dataset. Runs fully on-device with no cloud dependencies using Ollama with Qwen3-8B, ChromaDB for vector storage, and a hybrid dense + BM25 retrieval pipeline over curated schema documentation. Researchers ask natural language questions about complex event structures, get working PySpark code, and receive schema-aware warnings when field names don't match the actual data model. To select the best model, I benchmarked 14 models across 6,300 responses using a custom LLM-as-judge evaluation pipeline built on Prometheus 2, producing a statistically significant performance spread (Friedman p<0.000001) that guided the final model decision. Also serves as a proof of concept for on-device SLM deployment in research environments where data governance prohibits sending institutional data to external APIs.
Consumer App
A full-featured consumer mobile application built from the ground up — handling everything from onboarding and authentication to real-time data sync and payments. Clean architecture, smooth transitions, and the kind of polish that makes users forget how much is happening under the hood.
POS App — Cafe & Worker Flow
A point-of-sale system designed around the reality of high-volume cafe operations — fast order entry, queue management, and role-based worker flows that keep the line moving. Built to feel instant on every tap.
POS App — Restaurant Flow
The restaurant-facing side of the POS platform — table management, split checks, kitchen ticket routing, and real-time order sync across devices. Cross-platform from a single codebase, with offline-first data and sub-second sync when connectivity returns.
Get In Touch
[CONTACT_MESSAGE_PLACEHOLDER — A brief message inviting visitors to reach out]