Verified Receipts & Architectures

Production AI Projects & Case Studies

Real code builds demonstrating latency audits, token cost controls, and robust evaluation setups. Clean architectures with full source access.

Interactive Pipeline

LangGraph state machine logic mapping planning nodes, worker groups, tool routers, and human-in-the-loop validation.

Benchmarks verified in LangSmith evals.

Showcase

Real-world systems with open source code.

Multi-step agent searching the web, executing code checks, verifying citations, and writing markdown reviews.

LATENCY3.2s avg

TOKENS4.2k avg

EVALS94% golden accuracy

LangGraphFastAPITavily Search API

Production FastAPI backend processing document uploads, semantic splitting, Pgvector indexation, and reranking.

LATENCY140ms retrieval

TOKENS500 tokens/query

EVALS98% citation faithfulness

pgvectorCohere RerankFastAPIPhoenix

Standardized Model Context Protocol server exposing local shell execution, DB queries, and safe file writes.

LATENCY42ms tool latency

TOKENSN/A

EVALS100% schema validation

MCP SDKTypeScriptNode.js

A cyclic customer tickets workflow with automated triage nodes, policy check blocks, and human review gates.

LATENCY900ms cycle time

TOKENS1.8k tokens/cycle

EVALS91% automation rate

LangGraphOpenAI APIsPostgreSQL