AI Analysis

CurvLogic uses AI at every stage of the optimization pipeline — broad search analysis, search-around refinement, candidate evaluation, and autonomous orchestration decisions. Multiple leading AI providers are supported.

Multi-Provider Support

CurvLogic integrates with multiple AI providers, giving you flexibility to choose the model that best fits your needs and budget:

OpenAI — GPT-4o and GPT-4o-mini
Anthropic — Claude Sonnet, Opus, and Haiku
Google Gemini — Gemini Pro and Flash
xAI / Grok — Grok models

Each analysis stage can be configured with a different provider and model. You provide your own API keys, so you control costs and data handling preferences directly with each provider.

Analysis Stages

1. Broad Search Analysis

After a broad search completes, AI receives the full optimization results and performs a deep analysis:

Distribution analysis across key metrics (profit, drawdown, Sharpe, trade count)
Parameter sensitivity classification (strong, moderate, weak impact)
Cluster detection — identifying distinct promising parameter regions
Portfolio context — how results relate to your existing live EAs
Symbol and timeframe recommendations for the next stage

The output includes proposed narrower parameter ranges for search-around optimization, along with an EA summary and configuration guide.

2. Search Around Analysis

Each search-around run is independently analyzed. The AI receives the focused optimization results and extracts the best specific parameter sets as deployment candidates. This is the stage where broad ranges become precise, actionable configurations.

3. Candidate Analysis

For each candidate backtest, AI performs a comprehensive deployment readiness evaluation:

Multi-level recommendation: Strong Launch, Launch, Conditional, Skip, or Reject
Confidence scoring: 0.0 to 1.0 numeric confidence
Detailed reasoning: Concise summary of the analysis rationale
Risk assessment: In-depth analysis of the candidate's risk profile, including drawdown characteristics, trade distribution, and robustness indicators

4. Orchestrator Intelligence

In autonomous mode, AI makes pipeline advancement decisions — which EAs to process next, when results warrant moving to the next stage, and whether candidates are ready for deployment consideration.

Customizable Analysis

All AI analysis prompts are configurable through the Settings interface. This means you can:

Adjust the criteria and thresholds AI uses for recommendations
Customize the analysis format and depth
Add domain-specific instructions (e.g., prefer low-drawdown strategies)
Modify portfolio consideration rules

No code changes are needed — all prompt customization happens through the web interface.

Safety Guardrails

Human-in-the-Loop: AI suggestions and deployment recommendations always require explicit user approval before any EA is deployed to live trading. The review workflow ensures no automated deployment happens without human confirmation. AI provides intelligence — you make the final call.

Data Privacy

Only optimization metrics and parameter values are sent to AI providers — never account credentials, personal information, or trading account details. Each AI provider has its own privacy policy and data handling practices. You choose which provider to use and manage your own API keys.

← Back to Home