Inspiration

Everyone hates waiting 20 minutes on hold to customer support only to repeat account details to three different agents. We wanted to make telecom support as fast and effortless as texting a friend on WhatsApp—powered by an AI that already knows your account history, devices, and plans.

What it does

OmniResolve AI is an intelligent customer care assistant on WhatsApp. When a subscriber messages in, it automatically recognizes their phone number, pulls their complete Customer 360 profile from Neo4j (plans, bills, network status, active tickets), and uses an AI agent to troubleshoot issues, explain bills, upgrade plans, or seamlessly escalate to human agents—in seconds.

How we built it

WhatsApp Channel: AWS End User Messaging Social + Amazon SNS + Amazon SQS for reliable, bidirectional messaging.
Customer 360 Knowledge Graph: Neo4j database mapping customers, accounts, SIMs, billing cycles, and network towers.
Agent Orchestrator: LangGraph state machine coordinating tool calls (account lookups, billing checks, ticket creation).
LLM Reasoning: High-speed LLM inference (via Groq / Amazon Bedrock) paired with vector knowledge base retrieval (GraphRAG).

Challenges we ran into

Async WhatsApp Ingestion: Handling multi-layered AWS SNS-to-SQS JSON webhook payloads while guaranteeing strict message deduplication.
Sub-Second Graph Retrieval: Structuring Cypher queries across dense Customer 360 graphs so the agent retrieves full context without latency lag.
Tone & Hallucination Guardrails: Keeping the AI grounded strictly in verified telecom account data while maintaining a warm, natural conversation.

Accomplishments that we're proud of

Built a zero-hold-time customer experience entirely over WhatsApp.
True Customer 360 memory: the agent remembers past tickets and active plans without asking the user once.
Resilient background worker architecture with automatic SQS polling and failure recovery.

What we learned

How to configure and integrate AWS End User Messaging Social with Meta's WhatsApp Business Platform.
Combining Graph Databases (Neo4j) with LangGraph creates far more contextual and reliable agents than pure vector search alone.

What's next for OmniResolve-AI

Voice Support: Voice notes input and audio responses on WhatsApp.
Proactive Notifications: Automated WhatsApp alerts when bills are due or tower outages affect a user's registered area.
Self-Service Actions: Direct one-click UPI/card bill payments directly inside the WhatsApp chat.