AI Intern (Techcom Life)

Key Accountabilities (1)

• Research & Experiment (GenAI/NLP): prompt engineering, few-shot, RAG, tool/function calling, safety/guardrails.
• Knowledge Data Pipeline: collect → clean → label → normalize knowledge (FAQ, SOPs, policies, product sheets, forms), define metadata/taxonomy.
• RAG Stack: document ingestion (Confluence, PDF/Doc, portals), chunking, embeddings, indexing (FAISS/pgvector/Chroma), evaluate Recall@k/MRR.
• Conversation Design: intents/slots, fallback/disambiguation, orchestration with LangChain/LlamaIndex/Semantic Kernel.
• Quality & Evaluation: build test sets, A/B prompts, track accuracy/coverage, hallucination rate, latency.
• Integration & APIs: work with REST/GraphQL, FastAPI; collaborate with FE/BE to ship internal PoC.
• Voicebot (optional): explore STT/TTS, basics of telephony (SIP), integrate with 3CX/FTI/AWS Connect for PoC.
• Security & Compliance: handle PII responsibly, anonymize logs, follow internal policy & Vietnam regulations.
• Docs & Demos: write concise technical notes, ops checklists, and prepare periodic demos.

Success Profile - Qualification and Experiences

Must-Have Qualifications

• 3rd–4th year student or fresh graduate in CS/IT/AI/Data (or equivalent).
• Solid fundamentals in ML/DL; basic knowledge of NLP/LLM (tokenization, embeddings, transformers).
• Proficient in Python, notebooks, Git; familiar with common AI libraries (Hugging Face, LangChain/LlamaIndex is a plus).
• Strong data mindset: Vietnamese text preprocessing (Unicode/diacritics), regex/basic NLP pipeline, and basic evaluation concepts (precision/recall; BLEU/ROUGE is a plus).
• Able to read technical English; clear communication in Vietnamese; proactive, fast learner, team player.

Nice-to-Have

• Hands-on with chatbot/RAG/voicebot, hackathon experience, or a portfolio/GitHub.
• Experience with Azure OpenAI/OpenAI/AWS Bedrock/Vertex AI; vector DBs (Pinecone/Weaviate/pgvector).
• API/FastAPI, basic SQL (PostgreSQL), NoSQL; Docker; CI/CD; experiment tracking (MLflow/DVC).
• Basic frontend for demos (React, Streamlit/Gradio).
• Domain familiarity with insurance/sales.