← Selected systems

CASE STUDY / 03

Conversational support engine

Resolve common support requests at volume while preserving a route for complex questions.

Based on the public résumé. Architecture is a conceptual walkthrough; client implementation details and private evaluation data are not published.

Responsibility

The résumé describes building a customer-support chatbot with Rasa and transformer models during the Cognizant role.

Architecture at a glance

  1. Customer request
  2. Intent routing
  3. Response or fallback
  4. Support handoff

Reported outcome

50K+ reported daily interactions

Interaction volume describes workload, not unique users or resolved conversations. The PDF separately reports an 85% resolution rate.

The counting unit, observation window, resolution definition and baseline are not published. Volume and resolution should not be combined into an inferred count of successful cases.

Read the source résumé (PDF)
Technical deep dive

Design decision

Use intent routing and fallbacks to keep common requests efficient and make unresolved conversations actionable.

Alternatives and tradeoffs

A generative-only assistant can handle flexible language, but needs grounding and escalation safeguards. Structured flows trade conversational flexibility for predictable handling.

Engineering takeaway

Evaluate handoff quality and repeated contacts as well as resolution rate. A completed conversation does not automatically mean the customer problem was solved.