DataRev Labs
The assessment tells you where you stand. These labs show how we build: interactive simulators that teach the actual architecture behind RAG, agent governance, and the protocol stack production agentic systems run on.
AI Training & Learning Path
6-module, 9.0-hour executive & technical program: from LLM Fundamentals and Vibe Coding to Agentic Workflows, Harness Engineering, GraphRAG, and Governance.
Attention Lab — Transformer Architecture
Interactive guided walkthrough of the Transformer architecture: pre-processing, encoder, decoder, next-token probability distribution, and hyperparameters (Temperature, Top-N, Top-P).
RAG & GraphRAG Simulator
Retrieval-augmented generation from chunking through reranking, with a GraphRAG mode compared side by side against traditional RAG. Includes the tool/agent harness and an MCP walkthrough with runnable examples.
Agent Governance Simulator
How an agent gets supervised in production: guardrails, traces, human escalation, and the checkpoints that separate a pilot from a system that can actually be audited.
The Agentic Stack
Reference map of the agentic ecosystem's protocols — MCP, A2A, A2UI, Skills and tools — with comparisons, use cases, and a resource library.
People Analytics Pipeline & Simulator
End-to-end interactive People Analytics simulator: OLTP source architecture, ETL pipeline with 4 identity rules, SCD Type 2, executive KPI dashboard, and causal models.
Structured Data & LLMs Lab
Patterns for querying large-scale structured databases with LLMs: Text-to-SQL, semantic layers (Cube/dbt), metadata RAG, and AST security guardrails.
SEO vs. GEO Simulator
Architecture simulator comparing traditional SEO (PageRank, backlinks) vs. GEO (Generative Engine Optimization for Perplexity, ChatGPT Search, and Gemini).
Agent Security Simulator
The attack surface of an agent in production: prompt injection, tool poisoning and exfiltration, with the defence layers that contain them and a full incident walkthrough.
Reference Architectures
How an agentic system is actually assembled: the agent loop, orchestration, and the open-versus-closed-source call at each layer of the stack, with each tool's real licence.
The labs run as standalone applications inside this site, under the same DataRev identity.