Frontend · Web Design
Menteng Racquet Club
Editorial single-page site for a private tennis club in Jakarta, with a scroll-driven GSAP animation system.
HTMLCSSJavaScriptGSAP
Seoul, South Korea
Oswaldo
Dylan
Software Engineer. I build software across AI, robotics, and the web. I care how the UI feels as much as how the system runs, and I ship things people can actually rely on.
Scroll
I like owning problems the whole way through, from a robot's control loop to the pixels a client sees.
Currently building AI and robotics software at Luxolis. Before that, RAG systems, LLM research, and full-stack products.
0+
Years building software
0
Companies and labs
0
Languages spoken
Sungkyunkwan University
B.S. Computer Science and Engineering
Aug 2022 - Jul 2026 (Expected) · GPA 4.24 / 4.5
- Merit scholarship, 6 semesters running
- Coursework: Deep Neural Networks, Machine Learning, Operating Systems, Computer Security
Languages
English · FluentIndonesian · NativeKorean · Intermediate
Recent Projects
Frontend · Web Design
KELANA
Minimalist bilingual lookbook site for a Jakarta streetwear brand, with hand-crafted motion.
HTMLCSSJavaScript
Fullstack · Industrial Software
ROBOT UI
A control panel for a Doosan robot arm: build motion plans, record waypoints by hand-guiding the arm, then run them live.
PythonFastAPIC++Alpine.jsWebSocket
Work History
Mar 2026 - Present
CurrentAI & Robotic Software Engineer
Luxolis · South Korea
- Built the whole welding automation stack, an Alpine.js UI on FastAPI with a C++ DRFL robot daemon, that drives a Doosan A0912 arm, a Raycus CFL laser, and a motorized turntable from one piece of software.
- Containerized our software into reproducible Docker services with automated build scripts, bundling MinIO, Redis, and SQL.
- Shipped a React Native app for Android and iOS with 3D spatial vision, checking package compatibility and running integration tests across the stack.
C++FastAPIROS2DockerReact NativeDRFLNov 2025 - Feb 2026
AI / Backend Engineer, RAG Pipelines
Saifly · United States (Remote)
- Took a compliance RAG pipeline from a zero baseline to recall@k 0.448 and MRR 0.382, using Docling chunking, pgvector retrieval, and an eval harness that tracked recall, precision, MRR, and NDCG.
- Built semantic chunking with clause-level embeddings and token-boundary limits, plus metadata-rich indexing so retrieval could actually be explained.
- Cut ingestion overhead with async batching, dedup, and incremental re-indexing.
PythonpgvectorDoclingRAGEmbeddingsSep 2025 - Dec 2025
Undergraduate Researcher
HLI Lab, Sungkyunkwan University · South Korea
- Ran cultural-bias evaluations across GPT, Llama, Qwen, Claude, and Gemini to see how their cultural priors and stereotype sensitivity differed.
- Proposed a value-based alignment framework that infers a model's values, corrects for bias, scores against World Values Survey data, and steers personas using Schwartz's value theory.
LLM EvalPythonPrompt EngineeringResearch
Tools & Tech
Languages
PythonC / C++JavaScriptTypeScriptDartSQL
Frameworks
Next.jsReactReact NativeNode.jsFlutterFastAPITailwind
AI / ML
RAG PipelinesTransformersEmbeddingsVector SearchLangChainDoclingOllamaClaude APICNN / SVMSHAP / LIME
Infra & Data
DockerAWSLinuxGitFirebaseMySQLRedisMinIOpgvector