Avenra-AI

Avenra - AI startup intelligence: analyze pitch decks, KPIs & market data to score performance, visualize metrics, and deliver clear, actionable recommendations for founders and investors.

Comment

Inspiration

We built Avenra-AI to speed up and standardize how founders and investors evaluate early-stage businesses. The team noticed that reviewing pitch decks, spreadsheets, and market data is time-consuming and inconsistent. By combining AI with simple dashboards and KPI normalization, Avenra aims to surface the most important strengths, risks, and opportunities in seconds.

What it does

  • Accepts uploads (pitch decks and data files) and extracts core information.
  • Normalizes KPIs and converts raw inputs into comparable metrics.
  • Uses AI scoring and prompt-driven analysis to evaluate decks, KPIs and market context.
  • Produces visual dashboards, charts, and an overall score with actionable recommendations.
  • Exposes server API routes so the UI can request targeted analyses (kpis, market, extract, scoring).

How we built it

  • Built with Next.js (app router), TypeScript and React for a modular, component-driven frontend.
  • Server-side API endpoints handle analysis, extraction, scoring and market insights.
  • A lightweight AI helper layer constructs prompts and manages model interactions.
  • Firebase is used for authentication and file storage.
  • Reusable UI components power uploads, insights, dashboards, navigation, and charts.
  • Styling is organized with global CSS and PostCSS.

Challenges we ran into

  • Handling inconsistent input formats (PDFs, PPTX, CSVs) and extracting structured data reliably.
  • Prompt engineering: designing prompts that yield consistent, structured outputs rather than free-form text.
  • Avoiding AI hallucinations and ensuring the analysis is verifiable and reproducible.
  • Managing latency and cost for model calls while keeping the UX responsive.
  • Coordinating environment variables and secrets (model API keys, Firebase config) for local development vs deployments.

Accomplishments that we're proud of

  • Implemented an end-to-end upload → extract → analyze pipeline that integrates file handling, KPI normalization, and model scoring.
  • Built interactive dashboards (KPI charts and market analysis) that make recommendations actionable and easy to interpret.
  • Modularized the codebase so new analysis endpoints and AI prompts can be added quickly.
  • Created reusable components that speed up building new pages (analysis, dashboard, simulation).

What we learned

  • Quality of input matters: robust parsing and normalization reduces downstream model errors.
  • Prompt design and controlled output schemas are critical when you need structured data from models.
  • Next.js app-router provides a clear separation of server routes and UI, which simplifies server-side AI calls.
  • Small, incremental validations (sanity checks on extracted KPIs) prevent cascading failures in scoring.

What's next for Avenra-AI

  • Improve extraction: add more robust parsing for PDFs/PPTX and richer entity extraction.
  • Add automated tests for API routes and prompt outputs (schema validation of AI responses).
  • Implement role-based access and user dashboards with Firebase auth.
  • Add CI/CD and staging deployment; monitor model usage and cost.
  • Expand analytics: cohort comparisons, benchmarking against industry datasets, and downloadable reports.

Built With

Share this project:

Updates