Banking App PFM & CX Training Datasets + LLM Fine-Tuning

About OpenTrain

OpenTrain aggregates data-labeling and AI-training jobs from across the industry into a single job board so contributors can find work without hopping between dozens of sites.

Creating an OpenTrain account is free and applying takes only a few minutes.

About AI Training Work

AI models learn from examples prepared and reviewed by people. Tasks include annotating conversations, rating responses, and structuring data so models behave accurately and safely.

This project focuses on conversational and function-calling annotation for a banking AI assistant — the human annotations directly shape how the assistant understands user intent and executes backend queries.

The Role

We are hiring intermediate-level annotators to label a financial conversation dataset used for LLM fine-tuning and policy evaluation.

You will produce structured JSON records for each sample, using AWS SageMaker as the labeling platform, and deliver annotations that include intents, function calls, assistant response labels, and SQL safety judgments.

What You'll Do

Annotate conversational text samples for an AI financial advisor, following detailed guidelines provided in the labeling UI.

• Classify user intent (examples: spending insight, budgeting, card support).
• Label transaction and merchant categories where applicable.
• Evaluate assistant responses for tone, completeness, and professionalism.
• Tag function calls that map to backend operations (get_user_summary, get_transactions, run_custom_sql).
• Produce JSON records containing user query, labeled intent, function name (if any), arguments, and assistant reply.
• Validate SQL queries: mark whether queries are safe, parameterized, and read-only.

Requirements

You must meet all of the role requirements listed below. Do not apply unless you can demonstrate the required experience or skills.

• Familiarity with personal finance terminology and common banking concepts.
• Previous experience labeling chatbot or conversational data.
• Comfort with function-calling and API-style structured data (mapping inputs to names and args).
• Basic SQL literacy sufficient to identify unsafe queries and determine whether queries are parameterized and read-only.

Tools, Data Types, and Labeling

Labeling will be done in AWS SageMaker. The dataset is text-only and annotations will include classification, function-calling tags, fine-tuning targets, QA items, and text-generation labels.

You will deliver structured JSON records for each annotated sample following the provided schema.

• Labeling software: AWS_SAGEMAKER.
• Data type: TEXT.
• Label types: CLASSIFICATION, FINE_TUNING, FUNCTION_CALLING, QUESTION_ANSWERING, TEXT_GENERATION.

Payment & Contract Details

This is a contract role paid as a fixed-price project. The fixed price for the assignment is USD 2,000.

Exact scope, acceptance criteria, and milestone or delivery expectations will be provided when you are hired; do not assume ongoing work beyond this project unless stated in the contract.

• Employment type: CONTRACTOR.
• Currency: USD. Payment type: FIXED_PRICE ($2,000).

How It Works

If selected you will receive onboarding materials and detailed annotation guidelines. Work is completed in the SageMaker labeling interface and submitted as structured JSON.

Quality will be monitored against gold-standard examples. Feedback may be provided and annotations may require edits before final acceptance.

• Work remotely from anywhere (worldwide).
• Follow project guidelines closely; accuracy and consistency are required for acceptance and payment.

Who Should Apply

This project is a good fit for people with hands-on experience labeling conversational data and a working knowledge of personal finance and basic SQL.

Ideal candidates are detail-oriented, comfortable mapping natural language to structured function calls, and able to evaluate whether SQL is safe and read-only.

• Intermediate experience level required.
• No specific language or country restrictions; worldwide applicants are welcome.