Production Stack

Built with best-in-class tools

🔮
Qdrant
Vector DB
🕸️
Neo4j
Graph DB
Redis
Semantic Cache
🌌
Gemini
Embeddings
🦙
Groq
Llama 3 Inference
🤖
LangGraph
Agent Loop
🚀
FastAPI
Backend
📊
Ragas
Evaluation

Core Architecture

What makes this FAANG-level

🌳Phase 2

AST-Based Chunking

Parses code using tree-sitter into functions, classes, and methods — not arbitrary character splits. Preserves semantic boundaries.

🕸️Phase 3

GraphRAG + Neo4j

Maps function call graphs so when Function A calls Function B across files, the system knows they're related — even in different repos.

🔍Phase 3

Hybrid Search (BM25 + Dense)

Combines keyword precision (BM25) with semantic understanding (Gemini 1536-dim embeddings) using Reciprocal Rank Fusion.

Phase 3

Redis Semantic Cache

Caches LLM responses by query similarity (cosine ≥ 0.92). Cuts API costs by ~70% and reduces latency from 3s → 100ms.

🤖Phase 4

LangGraph Self-Correction

A stateful multi-step agent that retrieves, grades context quality, and reruns with a better query if the first answer is poor.

📊Phase 4

Ragas Evaluation

Automated benchmarking for Faithfulness, Context Precision, and Answer Relevance. Real numbers on your resume.