Project Info
This project did not submit a demo video on Devpost.
Inspiration
Coming out of the COVID-19 pandemic, I noticed something unsettling: people had forgotten how to talk to each other. The isolation, combined with our already screen-saturated lives, had created a perfect storm. Eye contact became rare, group interactions turned tense, and many individuals — especially young adults — felt disconnected, not just from others, but from themselves. This wasn’t just anecdotal. Social anxiety rates spiked. Relationship formation slowed. Even birthrates fell in parallel with a global decline in interpersonal connection. I created SosheIQ to address that — to reawaken the lost art of conversation and human connection in a post-pandemic world.
What it does
SosheIQ is an AI-powered social skills trainer designed to help users improve real-world interpersonal communication. It simulates lifelike conversations — complete with social cues, nonverbal hints, and subtle emotional signals — and evaluates how users respond. Key features: Simulated AI conversations with context-based feedback Nonverbal cue recognition (e.g., facial expressions, posture hints) Visual prompts and avatars generated using Imagen Social cue analysis for empathy, timing, and appropriateness Progress tracking over time Think of it as a personal trainer for your social intelligence.
How we built it
We built SosheIQ using: Next.js for a fast, interactive front-end Gemini API to generate dynamic, scenario-based dialogues and evaluate responses Imagen API to create expressive avatars and nonverbal visual cues Node.js backend for session tracking and scoring logic Google Cloud Run to host and scale our app effortlessly Tailwind CSS for a clean and accessible UI The architecture is modular to support future expansion into voice/video simulation.
Challenges we ran into
Nuancing human behavior: Teaching AI to recognize or simulate tone, intent, and nonverbal cues required thoughtful prompting and iterative fine-tuning. Realistic visuals: Integrating Imagen in a way that enhanced the experience without uncanny valley effects took UI/UX iteration. Feedback tone: Designing feedback that was honest but encouraging — like a real social coach — was harder than expected. Post-pandemic realism: Balancing realistic social scenarios with optimism (without making them too clinical or awkward) was a creative challenge.
Accomplishments we're proud of
Created a working AI model that gives real-time conversational feedback Integrated dynamic visuals that reflect tone and social context using Imagen Developed a design system that feels approachable and non-judgmental Built a scalable backend for tracking emotional/social growth Got validation from early testers who said it “felt like therapy but more fun”
What we learned
Social intelligence is just as trainable as technical skill Visuals can amplify empathy when used with intention People are eager for judgment-free spaces to practice being human again Crafting emotional intelligence into an AI product is not only possible — it’s essential Conversation is a two-way rhythm, not just scripted input/output
What's next
Integrate voice analysis and facial emotion detection for richer simulations Introduce customizable personas (e.g., date night, job interview, family dinner) Partner with therapists and educators to fine-tune simulation content Expand into HR and soft-skills training environments Build out a mobile version for on-the-go practice Our ultimate goal is to make SosheIQ a core part of how people build confidence in the one skill we all need — human connection.
This project did not link a GitHub repository.
Analysis
No indexed repository for this project, so there are no commit stats to show.
Technology
- Google GeminiUnchecked
- Next.jsUnchecked
- ReactUnchecked
No repository was indexed for this project, so these Devpost claims have not been checked against code.
AI coding agents
No repository was indexed, so agent usage could not be checked.
Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.
Codebase size
No repository was indexed, so there is no codebase to measure.
This project did not link a GitHub repository, so there is nothing to diagram.
This project did not link a GitHub repository, so its feature claims have not been checked against code.
Export this project's context (description, README, evidence, key source files) to chat with an AI agent elsewhere.