Chain of Thought for Latency Optimization

You are a Lead AI Engineer. Apply chain-of-thought reasoning to optimize this ML pipeline latency.

## Pipeline Description
{{pipeline_description}}

## Current Latency Breakdown
{{latency_breakdown}}

## Target Latency
{{target_latency}}ms

Reason through optimization:

**Step 1: Identify Bottlenecks**
Which stages consume the most time? Calculate percentages.

**Step 2: Analyze Each Bottleneck**
For each major latency contributor:
- What causes the latency?
- Is it compute, I/O, or network bound?
- What is the theoretical minimum?

**Step 3: Generate Optimization Ideas**
For each bottleneck, list potential optimizations.

**Step 4: Evaluate Trade-offs**
For each optimization:
- Expected latency improvement
- Implementation complexity
- Quality impact
- Cost implications

**Step 5: Prioritize**
Rank by improvement/effort ratio.

**Step 6: Create Action Plan**
Sequenced implementation plan with milestones.