Bloom

Let AI analyze your reviews and inventory. Get marketing insights and inventory analysis. Turn feedback into faster growth, plus AI-powered inventory analysis.

Comment

Inspiration

Businesses often struggle to efficiently process and act upon customer feedback and inventory scattered across various platforms. We were inspired by the potential of AI to transform this raw feedback from a time-consuming chore into a powerful engine for growth, helping businesses understand their strengths, fix weaknesses, and market more effectively.

What it does

  • Bloom is an AI-powered platform designed to help businesses leverage customer feedback. It automatically:
  • Analyzes reviews: Uses AI (Gemini) to summarize text reviews, identifying key positive and negative points.
  • Generates marketing ideas: Creates ready-to-use recommendations content (titles, text) based on positive customer feedback.
  • Provides inventory insights: Analyzes specific product/snack ratings (collected via targeted feedback) to generate AI recommendations on which items to prioritize, review, or replace, optimizing stock based on customer satisfaction.
  • Creates audio testimonials: Converts text reviews into natural-sounding audio testimonials using AI voice generation (ElevenLabs), adding emotional impact.
  • Monitors key inventory: Integrates with a camera and Raspberry Pi using Computer Vision to count stock levels of high-demand or frequently out-of-stock items (identified by review analysis).
  • Nota: No se incluyo un backup de la base de datos, ya que se tienen los modelos de las tablas dentro del proyecto. ## How we built it
  • Frontend: Vue.js with Tailwind CSS for a modern, responsive UI. Pinia for state management, and we leverage the vue-bits page for component usage..
  • Backend: Node.js with Express for building the REST API, handling business logic and AI integration.
  • Database: MySQL managed with Sequelize ORM to store user, business, review, and generated AI data. AI Integration:
  • Google Vertex AI: Using Gemini models (gemini-2.0-flash-lite) for text analysis and recommendation generation.
  • ElevenLabs API: For generating realistic text-to-speech audio testimonials with emotional nuance based on review ratings.
  • Scheduling: node-cron for automating weekly insight generation and periodic audio processing.
  • Computer vision: Raspberry Pi running local CV models (Gemini) connected to a camera, communicating results back to the main API. ## Challenges we ran into
  • API quotas & availability: Encountered strict rate limits with models and errors due to incorrect or unavailable AI model identifiers/versions in specific regions. This required implementing workarounds like delays or switching models.
  • AI Output Parsing: Ensuring the AI consistently returned data in the expected JSON format, requiring robust parsing logic to handle variations.
  • Debugging Integrations: Tracing issues across the full stack (Frontend -> Backend -> Database -> External AI APIs -> Cloud Storage) required careful logging and testing at each step.
  • Real-time & scheduled tasks: Balancing immediate feedback (saving reviews) with computationally intensive AI tasks (generating audio/insights) led to using background scheduled jobs. ## Accomplishments that we're proud of
  • Successfully integrating multiple distinct AI services (Gemini for text, ElevenLabs for audio) into a cohesive backend.
  • Building a functional full-stack application connecting a Vue frontend to a Node.js backend with a relational database.
  • Implementing a robust authentication system using JWT.
  • Designing and implementing automated background tasks for generating valuable insights (inventory recommendations, marketing ideas, audio generation).
  • Developing a practical concept for integrating real-world Computer Vision data with customer feedback analysis. ## What we learned
  • Full-stack web development with Vue and Node.js/Express.
  • Practical integration of various Generative AI APIs (Vertex AI, ElevenLabs) and handling their specific requirements and limitations (prompts, output parsing, quotas).
  • Database design and management using Sequelize ORM with MySQL.
  • Implementing secure user authentication and session management with JWT.
  • The importance of asynchronous processing and background tasks for performance.
  • Debugging complex, multi-system applications.
  • Fundamentals of Computer Vision application design for inventory management. ## What's next for Bloom This project has been a huge challenge; we've laid a solid foundation, but the growth potential is immense. We envision Bloom becoming an indispensable partner for companies looking to not just listen to, but thrive on, their customers' feedback.

Here's our journey:

  • We'd be excited to complete the development of the Computer Vision module. Imagine Bloom physically monitoring your key products, alerting you before popular items run out, and validating customer feedback on availability in real time.
  • Refining the AI ​​Advantage: We will continue to refine our AI prompts, leveraging Gemini and ElevenLabs to deliver even more accurate summaries, better decisions, and even compelling marketing approaches and audio testimonials that truly resonate with emotions.
  • Comprehensive Dashboard: We will transform the information into a dynamic analytics dashboard. Visualize trends, compare review sentiment with inventory levels, and monitor the impact of AI-powered marketing campaigns, providing businesses with clear, actionable insights at a glance.
  • Image Generation: We'll overcome quota hurdles and bring back AI-powered promotional graphics, delivering the full marketing package (text, voice, and images) directly from customer reviews.

Built With

Share this project:

Updates