AI-Driven Dental Interview Platform with Real-Time Avatar Interviewer -- 2
Budget: $25 – $50 USD
Project Overview
I am developing a full generative AI interview platform designed specifically for dental graduates entering clinical practice.
The platform conducts realistic, spoken job interviews using an AI interviewer (avatar) and provides real-time and post-interview coaching feedback based on employer expectations in dentistry. The aim is to provide feedback to user son their communication effectiveness and their formulation of responses – it is not to coach them on “what” to say.
The platform simulates real dental job interviews and provides structured, employer-lens feedback based on:
• clinical judgement
• articulation of responses (STAR technique and behavioural based interview techniques)
• risk awareness
• communication & consent
• verbal and non-verbal communication
• professionalism
This is not a generic AI avatar demo or HR chatbot.
The avatar is the delivery layer, not the intelligence.
Core Objective
To simulate a real dental job interview, where the AI interviewer:
• asks scenario-based clinical and professional questions
• responds dynamically to the candidate’s answers
• evaluates responses using a defined scoring rubric
• delivers structured, employer-lens feedback
Avatar Interview Requirements (Critical)
AI Interviewer (Avatar)
• Real-time spoken interview (voice in / voice out)
• Natural conversational pacing (interruptions included)
• Neutral, professional interviewer tone (not friendly chatbot)
• Configurable interviewer “style” (corporate / private / mentoring-focused/public)
Avatar tech may include:
• D-ID, Synthesia, HeyGen, Soul Machines, Tavus, or equivalent
• Or custom avatar + TTS solution
Important:
Avatar realism is just as important to interview logic and feedback accuracy.
Speech & Conversation Pipeline
Required flow:
1. Avatar asks interview question (TTS)
2. Candidate responds verbally (STT)
3. Response sent to evaluation engine
4. Evaluation engine:
o LLM scoring
o deterministic phrase-level red-flag detection
5. Avatar responds back
6. Feedback stored
7. Interview continues
8. End-of-interview summary generated on platform and by email.
Latency should feel conversational but does not need to be sub-second.
Interview Intelligence (Non-Negotiable)
Each answer must be evaluated against predefined dimensions:
Additional logic:
• Phrase-level red flag detection
• Emotion detection / facial analysis
• Rule-based score modifiers
• Triggered coaching scripts
The system must not rely on:
• confidence detection alone
• sentiment analysis alone
• generic “AI advice”
All scoring must be explainable.
Feedback Output (User Experience)
After Interview
• Clear, structured feedback per question on platform and by email:
o Strengths
o Areas for improvement (e.g. tone, pace, depth of answers, use of “umm” & “ah”, lack of application of STAR technique.
o Suggested rewrite
o “What the employer hears”
End-of-Interview Report - examples
• Overall score based on metrics provided by me (score out of 100)
• Employer Trust Score (0–10)
• Top improvement priorities
Technical Architecture (Avatar-First)
Frontend
• Web-based (desktop-first)
• React / Next.js preferred
• Webcam (audio required)
Backend
• Node.js or Python
• Session-based interview state management
• Secure audio handling
AI Stack
• LLM: OpenAI / Anthropic
• Speech-to-text: Whisper / Deepgram / AssemblyAI
• Text-to-speech: ElevenLabs / PlayHT / Azure TTS
• Avatar rendering: third-party API or SDK
Key Requirement
Interview logic, scoring, and feedback must be decoupled from the avatar layer.
This allows:
• avatar replacement later
• text-based fallback
• faster iteration
Content & Logic Ownership
I will provide:
• full question bank
• scoring rubrics
• phrase-level red flag dictionary
• gold-standard answers
• exact feedback scripts
Developers are not expected to design interview content.
Admin & Iteration Requirements
The system must allow:
• adding/editing questions without code changes
• updating phrase lists and scoring weights
• editing feedback text
• log-in page with options to choose from prior to interview (question types, avatar, random questions, etc)
Admin panel or config-based system required.
Out of Scope
• Generic career coaching
• Clinical decision support
Ideal Developer / Team
Looking for:
• Full-stack developer or small team
• Experience with:
o AI avatars
o real-time audio pipelines
o LLM evaluation workflows
o UX focus with high quality design elements
• Comfortable building structured AI systems, not demos
Not required:
• ML model training
• healthcare background or knowledge
I am developing a full generative AI interview platform designed specifically for dental graduates entering clinical practice.
The platform conducts realistic, spoken job interviews using an AI interviewer (avatar) and provides real-time and post-interview coaching feedback based on employer expectations in dentistry. The aim is to provide feedback to user son their communication effectiveness and their formulation of responses – it is not to coach them on “what” to say.
The platform simulates real dental job interviews and provides structured, employer-lens feedback based on:
• clinical judgement
• articulation of responses (STAR technique and behavioural based interview techniques)
• risk awareness
• communication & consent
• verbal and non-verbal communication
• professionalism
This is not a generic AI avatar demo or HR chatbot.
The avatar is the delivery layer, not the intelligence.
Core Objective
To simulate a real dental job interview, where the AI interviewer:
• asks scenario-based clinical and professional questions
• responds dynamically to the candidate’s answers
• evaluates responses using a defined scoring rubric
• delivers structured, employer-lens feedback
Avatar Interview Requirements (Critical)
AI Interviewer (Avatar)
• Real-time spoken interview (voice in / voice out)
• Natural conversational pacing (interruptions included)
• Neutral, professional interviewer tone (not friendly chatbot)
• Configurable interviewer “style” (corporate / private / mentoring-focused/public)
Avatar tech may include:
• D-ID, Synthesia, HeyGen, Soul Machines, Tavus, or equivalent
• Or custom avatar + TTS solution
Important:
Avatar realism is just as important to interview logic and feedback accuracy.
Speech & Conversation Pipeline
Required flow:
1. Avatar asks interview question (TTS)
2. Candidate responds verbally (STT)
3. Response sent to evaluation engine
4. Evaluation engine:
o LLM scoring
o deterministic phrase-level red-flag detection
5. Avatar responds back
6. Feedback stored
7. Interview continues
8. End-of-interview summary generated on platform and by email.
Latency should feel conversational but does not need to be sub-second.
Interview Intelligence (Non-Negotiable)
Each answer must be evaluated against predefined dimensions:
Additional logic:
• Phrase-level red flag detection
• Emotion detection / facial analysis
• Rule-based score modifiers
• Triggered coaching scripts
The system must not rely on:
• confidence detection alone
• sentiment analysis alone
• generic “AI advice”
All scoring must be explainable.
Feedback Output (User Experience)
After Interview
• Clear, structured feedback per question on platform and by email:
o Strengths
o Areas for improvement (e.g. tone, pace, depth of answers, use of “umm” & “ah”, lack of application of STAR technique.
o Suggested rewrite
o “What the employer hears”
End-of-Interview Report - examples
• Overall score based on metrics provided by me (score out of 100)
• Employer Trust Score (0–10)
• Top improvement priorities
Technical Architecture (Avatar-First)
Frontend
• Web-based (desktop-first)
• React / Next.js preferred
• Webcam (audio required)
Backend
• Node.js or Python
• Session-based interview state management
• Secure audio handling
AI Stack
• LLM: OpenAI / Anthropic
• Speech-to-text: Whisper / Deepgram / AssemblyAI
• Text-to-speech: ElevenLabs / PlayHT / Azure TTS
• Avatar rendering: third-party API or SDK
Key Requirement
Interview logic, scoring, and feedback must be decoupled from the avatar layer.
This allows:
• avatar replacement later
• text-based fallback
• faster iteration
Content & Logic Ownership
I will provide:
• full question bank
• scoring rubrics
• phrase-level red flag dictionary
• gold-standard answers
• exact feedback scripts
Developers are not expected to design interview content.
Admin & Iteration Requirements
The system must allow:
• adding/editing questions without code changes
• updating phrase lists and scoring weights
• editing feedback text
• log-in page with options to choose from prior to interview (question types, avatar, random questions, etc)
Admin panel or config-based system required.
Out of Scope
• Generic career coaching
• Clinical decision support
Ideal Developer / Team
Looking for:
• Full-stack developer or small team
• Experience with:
o AI avatars
o real-time audio pipelines
o LLM evaluation workflows
o UX focus with high quality design elements
• Comfortable building structured AI systems, not demos
Not required:
• ML model training
• healthcare background or knowledge