HRM Genesis
Sapient Model Studio
Create a task-specific model from a few examples. A few examples. A model built for your task.
Banking Support Intent Recognition
Understand what the customer is really asking for
A banking support message could be about transfers, fees, card issues, or many other detailed requests. This shows how the system learns to tell apart 77 types of banking support intents.
Agent task replay
How the agent creates the model
Result Comparison
See if it actually gets the customer request right
Test samples show what changed before and after training, especially whether the system stops mixing up similar requests like cash withdrawal fees and cash withdrawals.
Key effect: Key metric comparison
The fine-tuned model learns the task boundary from the provided examples and fixes common label confusions.
Fine-tuned macro_f1
Primary metric · n=3076
Base macro_f1
Zero-shot HRM-XL-1.2B
Fine-tuned Accuracy
Verified prediction accuracy
OOV Predictions
Closed label compliance
Test set probe: Test-set sample probe
More Showcases
More model creation scenarios
Different business tasks, one creation loop: describe the task, provide examples, get a model.
Customer Review Analysis
What did the customer praise, and what did they complain about?
Customer feedback is often not just good or bad. It may mention the app, support, pricing, delivery, and more in one comment. This shows how the system pulls out each point and understands the sentiment behind it.
[{"aspect": "purchase-booking-experience.ease-of-use", "sentiment": "positive"}]Enterprise Tool Calling
Can it turn an employee request into the right action?
If an employee says, “Create a high-priority ticket for this customer,” the system needs to choose the right tool and fill in the right details. This shows how a plain request becomes a concrete action.
[{"name": "places_list_by_radius_nearby_search", "arguments": {"lat": 118.2917, "radius": 1.0739, "lon": 129.2438, "src_attr": "https://www.example.com/wiki/16817684192", "kinds": "list"}}, {"name": "query_for_city_names_by_state", "arguments": {"state": "Texas"}}]Vehicle Owner Feedback Classification
Is this driver praising, complaining, or asking a question?
Driver feedback may praise a feature, complain about a problem, or simply ask a question. This shows how the system sorts feedback into three groups: praise, complaint, and inquiry.
inquiry
Natural Language Data Query
Can business users ask questions without writing SQL?
A business user might ask, “How many orders did each product line have last month?” without wanting to write SQL. This shows how the system starts from 10 SQL examples and learns to turn questions into data queries.
SELECT DISTINCT T1.city FROM addresses T1 JOIN people_addresses T2 ON T1.address_id = T2.address_id
How It Works
The agent handles model creation
Users confirm the task definition and plan without touching training parameters.
Describe Your Task
Tell the agent what model you want to create. It will infer the task type and output schema.
Provide A Few Examples
Upload a small dataset or use sample data to run the creation flow.
Get Your Model
Review test-set behavior and save a reusable model for the current task.
Private alpha
Create your first exclusive model
Users with an invitation code can proceed to the Agent guidance process. Those without an invitation code can still submit their requirements and join the Waitlist.