Prism

Inspiration

Modern professional life demands confidence, awareness, and safety — often at the same time. Many professionals, especially women and early-career leaders, struggle with two parallel questions:

Am I presenting myself effectively?

Is this interaction safe and appropriate?

These uncertainties show up everywhere — conferences, networking events, workplace chats, LinkedIn messages, and even hybrid social environments. The idea behind PRISM was to build an AI copilot that supports both outward presence and inward protection in one unified experience.

Instead of treating confidence coaching and communication safety as separate problems, PRISM connects them. Clear presence reduces misinterpretation, and early detection of red flags improves confidence. That synergy inspired the project.

What it does

PRISM is an AI Copilot for Presence & Protection.

It has two core intelligence modules:

Presence AI – Confidence Projection

Analyzes photos or context descriptions

Provides feedback on professional presence

Offers coaching aligned with specific situations (presentations, networking, interviews)

Protection AI – Communication Safety

Detects red flags in messages, emails, or DMs

Identifies patterns like urgency pressure, boundary violations, or manipulative tone

Generates tiered boundary-setting responses users can adapt

Together, the system helps users show up confidently while staying aware of potential social risks.

How we built it

PRISM was built as a full-stack AI web application designed for rapid iteration and real-world usability:

Frontend built using modern React/Next.js patterns for fast UI responsiveness

TailwindCSS used for clean, accessible design

AI analysis powered by multimodal LLM APIs for both image and text understanding

API routes structured for modular intelligence services

Cloud deployment via Vercel for instant scalability

The architecture intentionally keeps user data privacy-first, minimizing storage and focusing on real-time analysis.

Challenges we ran into

The biggest challenges were:

Balancing accuracy with responsibility. Communication analysis is sensitive. We had to ensure outputs feel supportive rather than judgmental.

Designing meaningful AI prompts. Generic outputs reduce trust. Iterating on prompts to generate contextual, actionable insights took significant refinement.

Privacy considerations. Handling images and messages required a local-first mindset and clear safeguards.

Scope control. Many feature ideas emerged, but focusing on a strong MVP was essential for delivering a coherent product.