Daniel Panrong Dai

About

Consulting-background AI practitioner who led a production LLM pipeline on Snowflake Cortex to map hundreds of thousands of messy OCR/form-based contract fields into standardized Salesforce fields, including confidence scoring and an LLM-driven feedback loop. Strong focus on real-world constraints—token limits, cost control, and evaluation without ground truth—paired with frequent stakeholder-facing progress reporting.

Experience

Education

Key Strengths

• Led production deployment of an LLM system to map large volumes of heterogeneous contract forms to standardized Salesforce fields
• Deep practical understanding of LLM token limits, cost tradeoffs, and context-window performance issues (e.g., needle-in-haystack failure modes)
• Strong at reducing problem scope via data engineering and constrained/structured outputs to improve accuracy and control cost
• Builds evaluation approaches when ground truth is missing using stakeholder/user review and measurable human feedback loops
• Experienced consulting-style communication with non-technical stakeholders; manages expectations by clearly explaining AI limitations and tradeoffs