Skip to content
Search jobs inside your AI agent. Claude, ChatGPT, Claude Code, Codex, Gemini, Cursor and Grok
Connect

Mihup

NewBe an early applicant

Conversational Designer

Posted Updated

Description

What you'll do

You will own the conversation design for live products, from first research to what ships.

Research
  • Plan and run primary research: field studies, contextual inquiries, user interviews and usability tests of live and
  • prototype flows
  • Run secondary research: competitor teardowns and benchmarking of voice assistants, domain and policy
  • reading, call and chat log analysis
  • Turn findings into personas, journey maps, utterance libraries and clear design recommendations
Intents and understanding

  • Define intent taxonomies and entity models with the NLU/ML team, including sample utterances across
  • languages and code-mixed speech
  • Spot overlapping or ambiguous intents and design how the assistant tells them apart
  • Use real usage data to find gaps, retire weak intents and add new ones
Confidence thresholds and error handling

  • Decide what the assistant does at each confidence level: act, confirm, offer choices, reprompt or hand off
  • Set and tune confidence thresholds with the ML team, balancing speed against the cost of getting it wrong
  • Design repair strategies for misrecognition, silence, barge-in, noisy environments and out-of-scope requests
Context sharing

  • Design how context carries across turns, sessions, devices and channels, so users never have to repeat
  • themselves
  • Define slot filling, follow-up and correction behaviour ("no, the other one", "make it 7 instead")
  • Specify what context passes to other systems and to human agents at handoff
Complex conversational flows

  • Design multi-turn, branching flows for complex domains such as automotive, banking and customer service
  • Document flows, dialogue specs, prompts and edge cases clearly enough for engineering and QA to build and test
  • Write system prompts and guardrails for LLM-based assistants, and test how they behave
Multilingual design

  • Design experiences in English, Hindi and other Indian languages, including code-mixed speech like Hinglish
  • Adapt tone, phrasing and persona for each language rather than translating word for word
  • Work with linguists and annotators to keep responses natural and consistent across languages

What we're looking for

  • 3–5 years of experience in conversation design, voice UX, interaction design or UX research, with at least
  • some of it on voice or chat products
  • A degree in design from NID, IIT (IDC or a design programme), Srishti or a similar design school
  • Strong primary and secondary research skills, and a portfolio showing how research shaped your design
  • decisions
  • A working understanding of how NLU systems work: intents, entities, confidence scores and fallbacks
  • Proven experience designing complex, multi-turn flows, including error handling and context across turns
  • Fluency in English and Hindi; one or more other Indian languages is a strong plus
  • Clear writing and the ability to explain design decisions to engineers, data scientists and business teams
  • Comfort using data, such as logs, analytics and test results, to improve flows after launch
Nice to have

  • Experience in automotive, in-car voice, contact centre or banking products
  • Hands-on work with LLM-based assistants: prompting, evaluation and guardrails
  • Experience with conversation design tools such as Voiceflow, and prototyping tools such as Figma or ProtoPie
  • Familiarity with speech recognition and text-to-speech, including how they fail in noisy or accented speech
  • Experience mentoring junior designers or interns

Skills

Apply

See also

Design jobs by country — openings, pay and top skills →