Skip to main content

Jella

🎯 Conversational AI

"Hi! I'm Jella, the conversational AI. How can I assist you?"

Overview​

Jella is the conversational AI specialist, powered by Rasa NLU. Jella handles intent classification, dialogue management, and helps train and improve conversational AI systems across the platform.

Agent Profile​

AttributeValue
IDjella
RoleConversational AI
PlatformInternal
ModelRasa NLU
StatusOnline

Capabilities​

  • Intent Classification — Understands user intentions
  • Dialogue Management — Maintains conversation context
  • Entity Extraction — Pulls key data from messages
  • Conversation Flow Design — Builds dialogue trees
  • Training Data Generation — Creates NLU examples
  • NLU Pipeline Optimization — Improves model accuracy

Performance Stats​

8,521
Total Tasks
89%
Success Rate
0.8s
Avg Response

Intent Classification​

Jella classifies user messages into intents:

intents:
  - greet
  - loan_inquiry
  - check_status
  - document_upload
  - schedule_call
  - get_quote
  - complaint
  - goodbye

Entity Extraction​

Jella extracts key entities from conversations:

EntityExampleExtracted
loan_amount"I need a $500K loan"500000
property_type"for a single family home"single_family
location"in Austin, Texas"Austin, TX
timeline"within 2 weeks"14 days

Dialogue Flow Example​

flowchart TB
    Start[User Message] --> Classify[Classify Intent]
    Classify --> |loan_inquiry| Loan[Loan Flow]
    Classify --> |check_status| Status[Status Flow]
    Classify --> |greet| Greet[Greeting Response]
    
    Loan --> Amount{Has Amount?}
    Amount --> |Yes| Property[Ask Property Type]
    Amount --> |No| AskAmount[Ask Loan Amount]
    AskAmount --> Property
    Property --> Location[Ask Location]
    Location --> Summary[Provide Summary]

Training Data Format​

Jella uses Rasa NLU training format:

nlu:
- intent: loan_inquiry
  examples: |
    - I need a loan
    - Looking for financing
    - Can you help me get funded?
    - I want to buy an investment property
    - Need money for a rental
    
- intent: check_status
  examples: |
    - What's the status of my application?
    - Where is my loan?
    - Any updates?
    - Has my loan been approved?

Integration with Other Agents​

Jella routes conversations to specialists:

flowchart LR
    User[User Message] --> Jella
    Jella --> |Lending| M[M Concierge]
    Jella --> |Technical| Quinn[Quinn]
    Jella --> |General| Assist[Crella Assist]

NLU Pipeline​

pipeline:
  - name: WhitespaceTokenizer
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CountVectorsFeaturizer
  - name: CountVectorsFeaturizer
    analyzer: char_wb
    min_ngram: 1
    max_ngram: 4
  - name: DIETClassifier
    epochs: 100
  - name: EntitySynonymMapper
  - name: ResponseSelector
    epochs: 100
  • M — Lending conversations
  • Crella Assist — Platform guidance
  • Quinn — Complex queries