Banking Chatbot Data Labeler: PFM, Function Calls & SQL

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. Creating an OpenTrain account is free and gives you access to projects where contributors directly shape how modern AI systems behave.

Why Work In AI Training

AI training (data labeling) is the human side of building AI — people prepare and review examples that models learn from. This work is highly flexible, often remote, and accessible: many projects need attention to detail or domain familiarity rather than prior AI industry experience.

This project focuses on conversational banking data, so you'll be contributing to systems that help people manage money, understand spending, and interact securely with financial services.

The Role

We are hiring intermediate-level annotators to label a financial conversation dataset used to fine-tune a personal finance AI assistant. You will produce structured JSON records for each sample that include the user query, labeled intent, function name and arguments (when applicable), and the assistant reply.

• Label types: intent classification, transaction & merchant category classification, assistant response labeling (tone, completeness, professionalism).
• Function-call tagging: identify and tag calls such as get_user_summary, get_transactions, run_custom_sql.
• SQL validation: mark whether SQL queries are safe, parameterized, and read-only.

What You'll Do

Work through conversational text samples and produce complete, structured JSON annotations according to task guidelines. Each labeled record will capture intent, relevant transaction or merchant categories, any function call and its arguments, SQL safety judgments, and an evaluation of the assistant response.

• Classify user intent (e.g., spending insight, budgeting, card support).
• Tag transactions and merchant categories embedded in user messages.
• Mark function calls and populate JSON fields: function name and arguments.
• Evaluate assistant replies for tone, completeness, and professionalism.
• Validate SQL queries: flag unsafe, non-parameterized, or non-read-only queries.

Requirements

You must meet all of the following project requirements. These are taken directly from the role description and cannot be substituted.

• Familiarity with personal finance terminology and common PFM concepts.
• Experience labeling chatbot or conversational data (annotating intents, replies, or dialogue structure).
• Comfort with function-calling / API-style structured data and mapping user intents to function arguments.
• Basic SQL literacy sufficient to identify whether a query is safe, parameterized, and read-only.
• Experience level: Intermediate.

Logistics, Tools, and Compensation

This is a contract role performed remotely and open worldwide. Work is text-only annotation using AWS SageMaker (labelingSoftware: AWS_SAGEMAKER). The dataset and guidelines will be provided within the platform.

• Employment type: Contractor.
• Data type: TEXT. Labeling focus includes CLASSIFICATION, FINE_TUNING, FUNCTION_CALLING, QUESTION_ANSWERING, TEXT_GENERATION.
• Payment: Fixed-price project — USD 2000 (from the provided job data).
• No additional language or country restrictions specified.

How To Apply

Create or sign in to your free OpenTrain account to view the full project and apply. Applications typically require a short profile, examples of relevant annotation experience, and completion of any required qualification tasks or checks listed on the project page.

• Be prepared to demonstrate prior chat labeling experience or relevant finance domain familiarity.
• You will complete labeling inside AWS SageMaker per the project's annotation guide.