What you will do

- Build and improve product search: hybrid keyword + vector retrieval, query understanding (Thai/English, typos, synonyms), ranking and re-ranking.

- Build the conversational shopping assistant: RAG over catalog, promotions and store content; tool calling for stock, price and order lookups; safe fallbacks and human hand-off.

- Use behavioral signals from the tracking pipeline and personalization layer to rank results per customer.

- Own evaluation: offline relevance sets, LLM-as-judge and human review, guardrail tests, and online A/B tests with the BI analyst.

- Ship and operate services on GCP (Cloud Run, Vertex AI, BigQuery, vector search) with monitoring for latency, cost and quality.

- Improve product data where it limits search: attribute extraction and enrichment with LLMs.

- Document and share patterns so other CRC teams can reuse them.

What you bring

-  3–7 years in ML, search or backend engineering, with at least 1–2 years shipping LLM or search features to real users.

-  Strong Python; production APIs (FastAPI or similar); containers and CI/CD.

-  Hands-on with embeddings, vector databases, RAG, prompt and tool design, and at least one major LLM API.

-  Search fundamentals: BM25, learning-to-rank or re-rankers, relevance metrics (NDCG, MRR, recall).

-  Experience designing evaluations and reading A/B results honestly.

-  Works AI-first: agentic coding tools (e.g. Claude Code, Cursor, Copilot) are part of the daily workflow for building, testing and reviewing code — mandatory for this role.

-  Comfortable with Thai-language text processing, or eager to learn it fast.