SocialSim

💡 Inspiration

Many neurodivergent students, especially those with autism, struggle with real-world conversations, not because they lack intelligence, but because they lack safe spaces to practice social interaction. We were inspired to build a tool that provides those students with a supportive, judgment-free environment to build confidence and social skills, powered by AI that understands emotional nuance.

🚀 What it does

SocialSim is a real-time social conversation simulator that acts as a supportive coach. Users engage in realistic, scenario-based dialogues - like joining a group project or attending an interview - and receive adaptive feedback and voice interactions. The platform provides:

Emotionally-aware AI dialogue using Amazon Bedrock
Expressive speech via Amazon Polly
Real-time voice input through Amazon Transcribe
Personalized coaching feedback based on conversation quality

It’s not just a chatbot - SocialSim is a compassionate companion helping students prepare for real life.

🛠️ How we built it

We built SocialSim using:

Node.js & Express.js for the backend server, and React.js & TailwindCSS for frontend
Amazon Bedrock for emotionally adaptive AI dialogue
Amazon Polly to generate realistic, expressive voice output
Amazon Transcribe to analyze user speech in real time
Amazon DynamoDB for session storage and retrieval

We designed the system to be modular and extensible for different scenarios, voice settings, and accessibility needs.

🧗 Challenges we ran into

Fine-tuning AI responses to balance empathy with realism
Latency issues when chaining transcription, inference, and speech synthesis
Handling AWS permissions for S3 and Polly securely yet efficiently
Maintaining session state across real-time audio and message flows

🏆 Accomplishments that we're proud of

Created an emotionally intelligent AI simulator tailored to neurodivergent users
Enabled fully voice-driven interaction using Polly and Transcribe
Developed a feedback engine that analyzes conversation quality across 5 criteria
Delivered a seamless user experience despite real-time processing across multiple AWS services

📚 What we learned

How to design AI conversations with empathy for users with unique communication needs
How to integrate multiple AWS services into a unified, low-latency pipeline
The importance of structured feedback for building user trust and learning outcomes
Best practices for session management in conversational AI

🔮 What's next for SocialSim

Add multi-language support for international students
Train custom AI models for more personalized coaching
Expand the scenario library to cover workplace, healthcare, and relationship contexts
Launch a mobile-first version with push notifications and audio playback
Partner with educators and therapists to bring SocialSim into schools and clinics