Product Integration Engineer → ML Engineer
Building production-ready AI systems · Taiwan 🇹🇼
Currently completing the Inference.ai ML Engineering Mentorship Program — a 10-week intensive focused on LLMs, RAG, fine-tuning, and agentic systems.
| Project | Description | Stack |
|---|---|---|
| 🎓 Voice Learning Assistant | Voice-powered AI tutor with RAG knowledge base for interactive learning and quizzes | Claude API · faster-whisper · edge-tts · Chroma · Streamlit |
| 🧪 RAG Lab | Composable RAG experimental platform with Chunking, Embedding, Retriever, Reranker selection & RAGAS evaluation | RAGAS · BM25 + Hybrid Vector · Cross-Encoder · Ollama/OpenAI · Streamlit |
| 🗂️ Multimodal Data Pipeline | Web / OCR / ASR extraction pipeline with MinHash deduplication, language filtering & PII stripping | Gradio · trafilatura · pytesseract · faster-whisper · datasketch |
| 🎯 Job Search Assistant | Multi-agent system that parses resumes, evaluates job fit, and drafts tailored cover letters with security guardrails | Google ADK · FastMCP · Gemini · Pydantic |
| 📊 Vertex AI Expense Agent | Corporate expense compliance checking ReAct agent with Human-in-the-loop (HITL) interrupts & dashboard | GCP · Vertex AI Reasoning Engines · FastAPI · Gemini · Terraform |
Week 1 ✅ LLMs & Prompt Engineering
Week 2 ✅ Transformer Architecture & Training
Week 3 ✅ Pretraining Data Collection & Extraction (Multimodal Data Pipeline)
Week 4 ✅ Retrieval-Augmented Generation (Voice Learning Assistant / RAG Lab)
Week 5 ✅ Supervised Fine-Tuning (SFT)
Week 6 ✅ RLHF & Model Alignment
Week 7 ✅ RAG Evaluation (RAGAS / Composable RAG Lab)
Week 8 ✅ Local Deployment & CPU Inference (Whisper / Ollama / Minilm)
Week 9 ✅ Agentic Workflows (Vertex AI Expense Agent with HITL)
Week 10 ✅ Multi-Agent Systems & Capstone (Job Search Assistant with ADK + FastMCP)
