DevVerse

DevVerse: AI-Powered Virtual Development Pod
"Bringing intelligence to requirements — and speed to solutions."
DevVerse is an AI-powered assistant that automates the transformation of plain English RFPs (Request for Proposals) into structured, implementation-ready development assets — all through an intelligent pipeline of role-based autonomous agents and Retrieval-Augmented Generation (RAG).
What is DevVerse?
DevVerse mimics a real-world project team using intelligent agents like:
🔹Business Analyst Agent – Extracts user stories
🔹Design Agent – Generates UI/UX components & architecture
🔹Developer Agent – Produces backend code & database structure
🔹Tester Agent – Builds test cases based on user stories and code
These agents collaborate to convert an RFP into project artifacts — automatically, intelligently, and instantly.
Problem We Solve:-
RFPs are often written in unstructured, plain English.
Converting them into usable formats (user stories, UI layouts, test cases, etc.) takes time, manual effort, and domain expertise.
Scaling this process across domains like finance, e-commerce, or healthcare becomes inefficient.
Our Solution:-
DevVerse automates this workflow using a smart, modular pipeline:
🔹Embedding : Creation RFP is encoded into semantic vectors
🔹Vector Search (ChromaDB) : Retrieves domain-relevant templates
🔹Agent Collaboration : Agents extract and generate structured outputs
🔹LLM Response : Final context goes to LLM (Gemini-1.5-flash) to generate artifacts
Tech Stack:-
Frontend : Streamlit
Backend : Python, CrewAI
LLM : API Gemini-1.5-flash
Vector DB : ChromaDB
Embeddings : Sentence Transformers
Agent Framework : CrewAI, Prompt Chaining
Agents in Action:-
Each agent plays a defined role in the pipeline:
🔹Business Analyst Agent : Extracts domain keywords and requirements; Generates user stories based on RFP understanding
🔹Design Agent : Selects UI layouts from template pool; Builds hierarchy and architecture diagrams
🔹Developer Agent : Suggests database schema and backend module layout (Extendable to generate code skeletons)
🔹Tester Agent : Derives test cases from user stories and functions; Suggests automation test structure
System Architecture Overview:-
[User Input: RFP]
↓
[Vectorization: Sentence Embedding]
↓
[ChromaDB Search]
↓
[Relevant Format Retrieval]
↓
[Agent Collaboration]
↓
[Final Prompt Assembly]
↓
[LLM Generation]
↓
[User Stories | Design | Code | Test Cases]
Impact:-
🔹80%+ reduction in manual effort
🔹Rapid transition from RFP to assets
🔹Standardized documentation across projects
🔹Domain-agnostic: works for fintech, healthcare, e-commerce, etc.
Future Scope:-
Add agents like UI/UX Designer, Scrum Master, Security Expert
Generate backend code skeletons and API specs
Integrate with tools like JIRA, GitHub Projects, Figma

DevVerse is not just a tool — it's the future of smart project planning.

During the hackathon, our team successfully transformed DevVerse from a conceptual idea into a functional AI-powered prototype. We began by identifying the key bottleneck in traditional software development workflows — manual processing of RFPs. From there, we designed the overall system architecture, broke down responsibilities across agents, and implemented a working pipeline that integrates: 🔹 RFP vectorization using sentence embeddings 🔹 Retrieval from ChromaDB for format templates 🔹 Orchestrated prompts using CrewAI agents 🔹 Output generation through the Gemini-1.5-flash LLM 🔹 A Streamlit-based frontend for user interaction and results display Over the hackathon period, we developed and tested multiple real-world RFP scenarios (e.g., e-commerce), refined agent collaboration, and ensured seamless input-to-output flow. We also added a chatbot-style assistant for status updates and improved usability. Despite the short timeframe, we were able to build a robust, scalable, and modular version of DevVerse — showcasing our ability to solve a real-world automation problem using cutting-edge GenAI.

DevVerse

視頻

技術堆疊

描述

黑客松進展

籌資狀態

獲獎賽道

Shortlisted For In-Person Hack