• Build and iterate on LLM & AI-powered product features • Design, prototype, and ship features that use LLMs (e.g., content generation, summarization, classification, semantic search, assistants, recommendations). • Work with engineers to integrate LLMs into the product via APIs or internal services (RAG, tools/functions calling, workflows, pipelines). • Define evaluation strategies for LLM features (e.g., human-in-the-loop evaluation, rubrics, prompt experiments, offline/online metrics). • Continuously refine prompts, data pipelines, and system design based on user behavior, quality metrics, and product goals. • Own product analytics for data- & AI-powered features • Partner with product managers and designers to define success metrics (e.g., adoption, engagement, conversion, retention, quality, time-to-value). • Instrument new features: define events, ensure proper logging, and validate that data is correct and trustworthy. • Analyze funnels, cohorts, user journeys, and experiment results to understand drivers of behavior and outcomes. • Translate insights into clear recommendations that influence roadmaps, prioritization, and feature iteration. • Work with real-world transactional data (SQL & NoSQL) Explore, clean, and transform data from transactional (OLTP), analytical (OLAP), and event-based systems. • Work across SQL (e.g., Postgres, Snowflake) and NoSQL (e.g., Redis, document/Key-Value stores). • Design data assets and features that are usable by both analytics workflows and LLM/ML systems. • Define and track data quality metrics (completeness, consistency, timeliness, drift, schema changes). • Build checks, monitors, and alerts to detect data issues that can affect analytics or AI/LLM performance. • Work with data and engineering teams to diagnose root causes and drive changes that improve data quality over time. • Use core ML concepts (feature design, model evaluation, bias/variance, generalization) to reason about LLM and non-LLM approaches. • When appropriate, build and evaluate lighter-weight or traditional models (e.g., scoring, ranking, classification) to complement or replace LLM solutions.