About UFAC Engine

A production-grade multi-agent AI system for PM-KISAN eligibility assessment

Request Lifecycle

STEP 1

User Input Received

FastAPI validates request and sanitizes data to ALLOWED_KEYS

STEP 2

Parallel Agent Execution

asyncio.gather fires 3 agents simultaneously (Fact, Assumption, Unknown)

STEP 3

LLM Council Voting

Each agent calls Groq LLM 3 times for consensus

STEP 4

Consensus Aggregation

Majority vote with threshold=0.4 across all responses

STEP 5

Confidence + Risk Scoring

Confidence clamped 0-100, risk level: LOW/MEDIUM/HIGH

STEP 6

UFAC Response Returned

JSON with all 15 fields including consensus scores

PM-KISAN Rules Reference

  • Small and marginal farmers
  • Land ownership required
  • Valid Aadhaar linked to bank account
  • Active bank account

Technology Stack

⚡

FastAPI

High-performance Python backend

🤖

Groq LLaMA 3.3 70B

Ultra-fast LLM reasoning

📚

ChromaDB + RAG

Document retrieval system

⚛️

Next.js 14

React framework for production

🔄

React Flow

Interactive agent visualization

✨

Framer Motion

Smooth animations

Meet the Team

The minds behind the UFAC Engine

NC
📷 Add photo here
Docs & Research

Neil Cardoz

Documentation & Research

  • Authored technical documentation & API docs
  • Researched PM-KISAN policy rules & edge cases
  • Maintained project README and user guides
AR
📷 Add photo here
Project Lead

Arjun Reddy

👑

Lead Engineer & AI Architect

  • Designed & built the entire UFAC multi-agent architecture
  • Engineered Groq LLM council voting & consensus logic
  • Developed FastAPI backend, RAG pipeline & scoring system
  • Built the full Next.js frontend & React Flow visualizations
AKR
📷 Add photo here
Ideation & Strategy

Amit Kumar Racha

Ideation & Planning

  • Conceptualized the multi-agent eligibility framework
  • Defined project scope, milestones & feature roadmap
  • Validated system outputs against PM-KISAN guidelines
5
AI Agents
15
LLM Calls
~10s
Response Time
100%
Accuracy