While we prefer candidates who can work full U.S. hours, we can accommodate those who are available at least until 1:00 or 2:00 PM CT. Strong collaboration and availability during this window are essential to ensure alignment with the team. Preferred Skills: Strong proficiency in SQL, database design, and schema understanding Experience with LLMs for Text-to-SQL, such as Codex, Text-to-SQL T5, SQL-PaLM, or SQLCoder Experience with agent frameworks such as AutoGen, or LangGraph; knowledge of LLM orchestration, tool calling, memory management, and multi-agent coordination Hands-on with , Python, LangChain, Agentic AI , Snow flake, MongoDB, LLM, ChatGPT and AskAT&T Key Responsibilities: Design and develop Text-to-SQL models that accurately convert natural language inputs into syntactically correct and semantically meaningful SQL queries. Work with domain experts to create and annotate NL-SQL pairs for training and fine-tuning LLMs or encoder-decoder models. Build and optimize semantic parsing and query generation pipelines using transformers, LLM APIs (e.g., OpenAI, Claude), and vector databases. Evaluate model outputs for accuracy, query safety, and relevance using automated and manual validation approaches. Implement entity linking, schema linking, and query disambiguation techniques for handling complex database structures and user intents. Collaborate with prompt engineers to integrate LLM-based query generation with contextual prompt templates and fallback logic. Develop tools to monitor and debug query generation issues, and build metrics dashboards to measure model accuracy, latency, and usage. Stay updated with advancements in semantic parsing, weak supervision, and hybrid symbolic/neural models for database querying. Design and implement agentic workflows where LLM-based agents plan, reason, and interact with tools to fulfill multi-step user queries RTH-Y - IBMFG2JP00008325