hey Contra, first post here.

I'm Nouman. I build AI agents and automations, and for the last while I've been deep in the boring side of AI that nobody posts about: compliance.

most teams I talk to are shipping AI features fast and have no idea if they'd pass an EU AI Act check. usually they wouldn't, and usually it's fixable in a few weeks, not months.

so that's what I work on. agents and automations that actually run in production, plus the logging, risk checks and evidence trail that keep them legal.

I'll be sharing what I learn here, mostly practical stuff. if you're building with AI and have a question about either side, ask away.

what's everyone here building right now?

The evidence trail you mentioned is the hard engineering part. What do you retain for each agent tool call so someone can reconstruct the run without storing sensitive payloads?

Data Product Builder Agent

Technology & Tools: Generative AI, Agentic AI, AI Agents, Multi-Agent Systems, LLM, RAG, MCP, Vector Databases, Embeddings, Semantic Search, Hybrid Search, LangChain, LangGraph, CrewAI, AutoGen, OpenAI, Claude, AWS Bedrock, Cursor, Python, FastAPI, REST APIs, Docker, Git, GitHub, AWS, S3, Lambda, OpenSearch, PostgreSQL, CI/CD, Prompt Engineering, Context Engineering, Tool Calling, Function Calling, Structured Outputs, Human-in-the-Loop, AI Guardrails, LLMOps Role: Data Engineer / AI Engineer • Designed an enterprise-grade AI-powered Data Product Builder Agent to accelerate creation, validation, documentation, governance, and management of reusable data products. • Implemented Agentic AI and Multi-Agent Systems with specialized agents for metadata discovery, schema analysis, transformation, validation, documentation, orchestration, and deployment. • Built stateful LLM workflows using LangChain and LangGraph for reasoning, tool calling, structured outputs, agent memory, error handling, and Human-in-the-Loop approvals. • Implemented RAG using embeddings, Vector Databases, semantic search, hybrid search, and knowledge bases to ground LLM responses in enterprise data and metadata. • Integrated MCP for standardized context and tool integration and used CrewAI and AutoGen patterns for multi-agent collaboration and task delegation. • Integrated OpenAI, Claude, and AWS Bedrock for model flexibility, prompt/context engineering, evaluation, guardrails, and LLMOps. • Developed Python and FastAPI REST APIs and integrated S3, Lambda, OpenSearch, and PostgreSQL for data, metadata, retrieval, and application workflows. • Used Cursor, Git, GitHub, Docker, and CI/CD for AI-assisted development, version control, containerization, testing, and deployment