AI/ML Engineer — Classification + Matching

We're building an AI system that reads content and automatically determines:
 

1. Whether it contains a paid endorsement (yes / no)
2. What type of endorsement it is, based on our classification spec
3. Which brand it refers to — matched against our existing brand database, with a suggested new brand name when there's no match
    

We already have a labeled set of examples across the different endorsement types that you can use for training and evaluation. 
 

What you'll do

• Build a classification pipeline that flags paid vs. non-paid endorsements and categorizes endorsement type per our specification
• Build brand matching (entity resolution) against our database using fuzzy and/or semantic matching, with a "suggest a new brand" fallback when nothing matches
• Use our labeled examples to train and/or tune the system, and set up an evaluation framework so we can measure and improve performance per task
• Process large volumes of entries efficiently and cost-effectively

• Document the approach and hand off a maintainable system

   

What you should have

• Strong experience building NLP / text-classification systems
• Hands-on work with modern LLMs (e.g., Claude, GPT) — prompt engineering, structured outputs, and/or fine-tuning
• Experience with entity resolution / record linkage — fuzzy or semantic matching, embeddings, vector search
• A rigorous, evaluation-driven approach — you measure performance properly rather than by feel
• Comfort with our stack: Python, SQL, working with APIs and a database