build · 2026 · Lead Full-Stack & ML · MVP shipped
ShweMyay (ရွှေမြေ)
AI-powered AgriTech — offline Plant Doctor for Myanmar farmers
Lead Full-Stack & Machine Learning Developer · ShweMyay · Myanmar AgriTech
Offline-first
Mode
MobileNetV2
MVP model
<15MB
Target size
FastAPI
API
Challenge
Rural farmers lack expert access and often have poor or zero internet — cloud-only plant diagnosis fails when it matters most.
Approach
- 01Embedded a lightweight Deep Learning model (MobileNetV2 MVP via TFLite) on-device for instant, offline disease diagnosis in Burmese.
- 02Designed Edge–Cloud hybrid: on-device for mobile; FastAPI + Base64 image fallback for web where TFLite isn’t available.
- 03Architected a two-tier pipeline (crop detector → disease specialist) targeting >95% accuracy under ~15MB app size — YOLOv8 classification path.
- 04Built Colab MLOps: Kaggle ingest, augment, split, high-epoch training with early stopping.
- 05Offline-first Hive cache for forum/market data with background sync; JWT + rate limiting on the API.
Outcomes
- ▸Zero-latency Plant Doctor that works without connectivity
- ▸Cross-platform parity via edge + server fallback
- ▸Senior signal: MLOps + mobile Edge AI + secure API in one product
ShweMyay (Golden Land) is not a demo camera filter — it’s Edge Computing for Myanmar agriculture: diagnose on the phone, advise in Burmese, sync when the network returns.
The MVP ships MobileNetV2 via TFLite; the architecture already plans YOLOv8 hierarchical specialists and a future VLM Tier-3 cloud agent for rare species.
Hiring managers see seniority in the glue: dataset pipelines, on-device constraints, FastAPI fallback, Hive offline sync, and security middleware — not just a model notebook.
Stack
Need Edge AI or AgriTech mobile? I can take dataset → TFLite → Flutter offline UX with the same hybrid architecture.