• 

  MongoDB Atlas Server

• 

  DAW

• 

  Kids

• 

  Solana

Wonder: The Cursor for Music

One-Sentence Summary: Wonder is a custom, AI-driven DAW—a music production environment built from the ground up with AI at the center, moving beyond simple protocol shims or legacy software plugins.

Inspiration

We named the project after Stevie Wonder—a tribute to a pioneer who used emerging technology to push the boundaries of what music could be. We built Wonder to solve the "Interface Friction" for the specific creators on our team:

• Zach: A guitar player who wants to focus on playing without the learning curve of a traditional DAW.
• Kori: An experienced producer who wants to eliminate "menu-diving" and stay in a creative flow state.
• Nike & Gavin: Creative thinkers who lack formal music school training but have massive ideas that need a streamlined path to becoming full tracks.

What It Does

Wonder is your primary workspace. Instead of mapping MIDI CCs or hunting through nested folders, you interact with your production environment in plain language.

• Native AI Core: This is not an MCP (Model Context Protocol) integration or a "chat box" bolted onto a legacy DAW. The AI is the core engine of the product, allowing for tighter hardware-to-model integration.
• Intelligent Library Management: Wonder performs a recursive scan of your physical disk to find sounds, ensuring that your library is always indexed and searchable by vibe or technical characteristics.
• On-Demand Generation: If your local library lacks a specific fit, Wonder uses ElevenLabs to generate the exact sound requested and instantly adds it to your local and cloud-synced session.
• Cloud-Synced Identity: Your session data and user profile are hosted via MongoDB Atlas, allowing your creative environment and custom tags to follow you across any machine.
• Direct Web3 Distribution: We integrated Solana to revolutionize how music is shared. Users can export their tracks as a Solana Blink, allowing music to be instantly playable and actionable across any platform that supports Solana Actions.

How We Built It

We engineered a high-performance pipeline that balances local audio speed with cloud intelligence:

• Gemini (Google Generative AI SDK): Our primary "listener" that performs multimodal analysis on raw audio to provide categorical labels, vibe-checks, and descriptive tags.
• MongoDB Atlas: The backbone for cloud storage, managing user profiles, metadata persistence, and session states.
• ElevenLabs: Integrated for real-time synthesis of missing samples directly into the production timeline.
• Solana Actions & Blinks: Beyond distribution, we used Solana as a high-speed settlement layer to offset AI token costs. By allowing users to sell their high-quality, ElevenLabs-generated sounds to other creators on the network, we created a circular economy that subsidizes the compute costs of the generative engine.
• LanceDB: Powers our Library Intelligence. We use LanceDB for high-speed sample comparison, utilizing a hybrid of semantic tags (AI-generated vibes) and raw audio data (mathematical descriptors).
• Snowflake: Serves as a critical developer tool for Session Intelligence. By ingesting granular session data, Snowflake allows us to analyze exactly which Wonder tools are being utilized and how the AI is responding.

Challenges We Ran Into

• The MCP Latency Barrier: We originally explored using a Model Context Protocol (MCP) approach, but quickly found that the protocol overhead created too much latency for a professional music environment. To achieve the real-time response required for live production tools, we had to pivot away from MCP and build a fully independent, custom DAW from the ground up.
• Stateful Representation: Maintaining a "stateful" conversation during a long session was a hurdle. We transitioned to the official Google Generative AI SDK and shifted to agentic models. This transition was critical; by treating the AI as a state-aware agent, we were able to store comprehensive session histories that the Snowflake API could then ingest and analyze for creative and technical insights.
• Latency Control: We had to meticulously tune our processing threads to ensure that recording, AI analysis, and playback did not stutter or create audible lag in the buffer.
• API Orchestration: We developed sophisticated batching and prompting strategies to manage API rate limits and quotas without interrupting the creative process mid-session.

Accomplishments That We're Proud Of

• Full-Stack Ownership: Building the DAW architecture from scratch allowed us to bypass the overhead of legacy software. This control enabled us to cut down processing latency by minutes, delivering speedy, immediate results that keep the user in the creative zone.
• Slang-to-Signal Processing: Our search engine successfully maps producer terms like "thumpy" or "airy" to actual frequency ranges and transient profiles using a blend of vector search and math.
• The Wonder Workflow Integration: We successfully implemented a system where generative sound effects from ElevenLabs are automatically stored to the user's cloud account and tagged by the AI. These new assets instantly join existing samples in the unified Wonder Workflow.

What we learned

We learned that building an AI-native DAW is primarily a challenge of State Management. Shifting from stateless API interactions to a fully Agentic Model allowed us to maintain a persistent creative context, which is essential for professional production.

By prioritizing Full-Stack Ownership, we learned that we could achieve immediate, low-latency results that are impossible when using generic protocol shims. This architecture proved that "Real-Time AI" in music depends on the model having a direct, state-aware connection to the DAW's core engine. Finally, we learned that a robust Data Feedback Loop via Snowflake is non-negotiable for scaling; it allowed us to turn raw session logs into actionable insights for model refinement and system optimization.

What's Next for Wonder

• Blockchain-Powered Asset Control: In the future, we could use Solana to give creators stronger protection and control over their music assets. By tying ownership and transaction records to the blockchain, creators could have clearer proof of ownership, more transparent tracking of how their music is shared, and a more direct way to manage peer-to-peer exchanges. This could reduce reliance on middlemen and help ensure creators keep more control, visibility, and value from their work.
• Enterprise-Level Scalability: Leveraging our Snowflake and MongoDB backbone to handle massive influxes of user data and session concurrency. We are optimizing our pipeline to support thousands of simultaneous AI-driven DAW sessions without performance degradation.
• Cloud-Based Collaborative Jams: Transforming Wonder into a collaborative space where multiple producers can contribute to the same session in real-time via our cloud-first infrastructure.
• Customer Outreach and Community Engagement: We plan to aggressively market Wonder within professional audio communities, targeting mixing subreddits and specialized Discord servers. Additionally, we will initiate pilot programs with music production students at Belmont and Vanderbilt to refine the tool for the next generation of industry professionals.

Built With

• elevenlabs
• gemini-api
• google-generative-ai-sdk
• google-text-embedding
• lancedb
• librosa
• mongodb-atlas
• python
• snowflake-api