SparkConnect AI

Inspiration

Our inspiration stems from a severe structural marketplace mismatch in Singapore's social sector. On the demand side, understaffed non-profits and small NGOs suffer from massive administrative overhead and lack the technical vocabulary to define their specific operational bottlenecks. On the supply side, there is a massive pool of highly skilled student initiatives and Co-Curricular Activities (CCAs) looking to complete their Values in Action (VIA) hours. However, legacy volunteering platforms rely on rigid keyword searches, forcing these talented youths into generic, low-impact tasks—like a media club standing at a donation booth instead of handling creative production. We realized that the volunteer gap isn't a lack of human capital, but a systemic breakdown in visibility and skill alignment.

What it does

SparkConnect AI is a two-sided marketplace mobile application that acts as a "Spotify Blend" for social impact, instantly matching student group skills with real-world non-profit needs. It targets the Theme C challenge through three distinct features:

AI Problem Identification: NGO coordinators simply drop an unstructured text or voice "brain dump" about their daily frustrations into the app. The AI reads between the lines, diagnoses the hidden technical bottleneck, and translates it into a structured, actionable "Diagnostic Project Flag."

Dynamic Volunteer Coordination: Students use a Master Context Toggle to switch their profile capacity (e.g., Solo/Grassroots, CCA, or Full Class). The app serves a Tinder-style swipe feed of matched projects perfectly scaled to their group size and specialized skills.

Upfront Impact Measurement: Before a match is even made, the platform quantifies the exact hours, manpower, or resources the student project will save the NGO, shifting the focus from checklist volunteering to deep social alignment.

How we built it

We focused on building a lean, highly visual mobile-responsive frontend to validate our core workflow within the sprint.

Frontend UI/UX: We used React/Next.js and Tailwind CSS (accelerated by v0.dev) to build the interactive mobile prototype, ensuring a frictionless user flow across both the NGO and student journeys.

AI-Assisted Build & Planning: We used Claude, Google Gemini, and NotebookLM throughout—to pressure-test the concept, structure the feature set, write the matching logic, and refine the product flow against the challenge rubric.

AI Engine (design): We designed the AI diagnostic flow around a gpt-4o-mini inference layer and prompt-engineered the semantic mapping logic—analyzing the qualitative paragraphs of NGO struggles, mapping them against student capabilities, and calculating an "Alignment Score." For this MVP, we implemented the inference as a designed-and-mocked layer so the full user flow runs flawlessly in the demo; wiring the live model is our immediate next build.

Data Architecture: We designed a relational schema to handle dynamic user types (Volunteer vs. NGO) and implemented localized mock states for the demo to ensure flawless execution without database timeouts.

Challenges we ran into

Our biggest challenge was avoiding "AI Fluff." We recognized early that letting an LLM blindly diagnose problems and arbitrarily assign student projects could produce disconnected or unhelpful results for charities. Our design answer is a "Human-in-the-Loop Safeguard": when the AI surfaces a structural diagnosis, it is presented to the NGO admin with clear [Approve & Add] or [Dismiss] controls, ensuring the AI acts as an assistant rather than an autonomous decision-maker. This principle shaped our entire product flow.

Accomplishments that we're proud of

We are proud of evolving a traditional "volunteering directory" concept into a venture-grade, two-sided marketplace. By identifying the core friction point—that non-profits do not have the time to draft perfect technical project briefs—we reframed the problem entirely. We are also proud of our Master Context Toggle, which elegantly lets a single student leader switch between organizing a 40-person class deployment or a 5-person specialized tech sprint without creating multiple accounts.

What we learned

We learned that true AI leverage in the social sector isn't about building a simple chatbot wrapper; it is about using AI as an "organizational diagnostician" to uncover hidden workflow bottlenecks. Diving into Singapore's community ecosystem also taught us the critical distinction between hyper-local neighborhood Grassroots/RCs (Residents' Committees) that need bite-sized micro-volunteering, and larger SSAs (Social Service Agencies) that need long-term technical collaborations.

What's next for SparkConnect AI

For our next phase, we want to support regular, long-term volunteering relationships, which the National Volunteer and Philanthropy Centre advocates for over one-off events.

Live Model Integration: Wiring the gpt-4o-mini diagnostic layer from mocked to live, closing the loop on real-time problem diagnosis.

System Integration: We plan to integrate with Singpass for seamless identity verification, and partner with schools to auto-sync completed project impact metrics directly with formal VIA hours.

Scale and Scope: We aim to collaborate with SG Cares Volunteer Centres to deploy SparkConnect AI as a digital tracking tool for town-level deployments.

Collaboration Workspaces: Expanding beyond our MVP's "Connect Now" match chat, we will build dedicated workspaces for role assignment, availability scheduling, and 3-to-6 month project blueprints.

Built With

• claude-(anthropic)
• google-gemini
• lucide-react
• next.js
• notebooklm
• react
• tailwind-css
• v0-by-vercel
• vercel