TDH LABS
Home / Case Study
Case study · AI trading infrastructure

Orchestrating market context, structured model analysis, and MT5 execution.

An engineering case study about system boundaries, not a claim that an LLM can predict markets reliably.

At a glance

Problem: Combine news, macro context, calendar events, and MT5 positions into one repeatable analysis pipeline.
Role: Workflow and execution architecture.
Stack: n8n, Claude/OpenAI, structured JSON, MQL5.
Status: Reference product/system.

Workflow

  1. Collect headlines, sentiment, economic-calendar events, and current positions.
  2. Normalize times and label events as past, imminent, or just released.
  3. Send bounded context to an LLM and require structured output.
  4. Validate symbols, actions, risk fields, and stale signals.
  5. Route approved instructions to a signal handler and MQL5 execution layer.

Critical safety boundaries

LLM outputs can be wrong, inconsistent, or fabricated. A production system needs deterministic validation, risk caps, allowlists, stale-signal rejection, human approval where appropriate, full logs, and a kill switch.

What this proves

The project demonstrates cross-system orchestration: scheduling, external data, time-aware transformations, structured AI output, validation, notification, and MT5 integration.

View product detailsDiscuss a similar system
Software engineering example only. Not financial advice. No guarantee of profitability or model reliability.