Phnom Penh, Cambodia
Eng Sothy
Backend & AI Engineer
Data Science & Engineering student at the Royal University of Phnom Penh who builds production-shaped backend APIs and integrates machine-learning models behind them. Comfortable across the full path from data and model to a deployed, containerized service. Seeking a backend or AI/ML engineering internship.
About
I like seeing a project through end-to-end — from training a model or designing a schema, to shipping it behind a clean, documented API. I care about getting the fundamentals right: sensible data models, tested code, and services that are containerized and easy to deploy, not just code that works on my machine. Outside of coursework, I pick up real-world-shaped problems on my own — visual search, NLP, and now real-time event-driven systems — because that's where I learn the most.
Technical Skills
Languages
Backend
AI / ML
Infra & Tools
Also
Projects
Auto-Parts Image Search Service
View on GitHub →- Built a visual search service: upload a photo of a car part and get ranked matching products from the catalog.
- Engineered a multi-stage ML pipeline — YOLOv8 part detection, PaddleOCR for brand/part-number text, CLIP image embeddings, and BGE-M3 text embeddings.
- Indexed embeddings in FAISS for fast nearest-neighbor retrieval and merged catalog + vector results with confidence-weighted scoring.
- Exposed the pipeline as a documented REST API and packaged it with Docker for deployment.
Garage Service Provider — Backend API
View on GitHub →- Designed and built the backend for a garage service platform: authentication, admin and technical-staff accounts, and the product/inventory domain.
- Implemented JWT login with role-based access control and a clean layered structure (models, repositories, routers, schemas).
- Delivered full CRUD for categories, products, and inventory, including stock restock/deduct and low-stock alert endpoints.
- Containerized with Docker Compose and added a pytest test suite (130+ commits).
AI Mood Journal
View on GitHub →- Cross-platform journaling app that analyzes entries to surface emotional patterns and the topics driving them.
- Trained NLP models for multi-label emotion classification and used LDA topic modeling to explain the “why” behind a mood.
- Built a Flutter mobile front end backed by a FastAPI service exposing the ML models as endpoints.
SLAP Autofix
In Progress- Built the backend for a real-time roadside-assistance system: a stranded driver raises an SOS, nearby mechanics are notified, bid on the job, and get matched to the request.
- Developed the service in Spring Boot (Java 21) with JPA/Hibernate entities (PostGIS geography types), Java record DTOs, and MapStruct mappers across a clean layered structure.
- Designed an event-driven flow over Apache Kafka (sos.created, sos.bid.created, keyed by sos_id) and streamed live SOS/bid updates to drivers and mechanics over WebSocket / STOMP.
- Implemented two-stage geospatial matching (H3 grid-disk shortlist → PostGIS ST_DWithin exact filter) to surface the nearest available mechanics fast.
- Research project with my professor; I own the SOS (Sign of Service) module. Currently in progress.
Certifications
The first three make up Andrew Ng's Machine Learning Specialization (DeepLearning.AI & Stanford University).
Education
Royal University of Phnom Penh
B.Sc. in Data Science and Engineering — in progress
Phnom Penh
ETEC Center
Computer Programming, C/C++
2023–2024
Hun Sen Chumpuvorn High School
Graduated with honors
2023
Languages
- English — Fluent
- Khmer — Native
Contact
Looking for a backend or AI/ML engineering internship. Feel free to reach out.