Build a Cloud Accelerator Procurement and Reservation Simulator — implementation

## Build
Simulate capacity procurement strategies across on-demand and reserved instances for open-model serving. Synthesize multi-cloud reservation proposals using hosted Claude.

## User
Infrastructure Capacity Planner

## Workflow
1. Ingest historical inference request traffic patterns and multi-cloud accelerator instance pricing sheets.
2. Model baseline and peak capacity demand curves across simulated 1-year and 3-year commitment tiers.
3. Prompt hosted Claude to critique procurement risk, spot preemption vulnerability, and cost optimization levers.
4. Assemble a comprehensive procurement proposal dossier with projected cash flow impact.

## Interface
Spreadsheet-style simulation UI with downloadable procurement commitment briefs.

## Data
Tabular CSV files containing normalized spot, on-demand, and reserved pricing for cloud GPU instances.
Starter examples below are synthetic. They are not downloaded records.

## Source
[AWS Neuron documentation](https://raw.githubusercontent.com/aws-neuron/aws-neuron-sdk/master/README.md)
Incorporates AWS accelerator instance deployment principles (such as Inf2 and Trn1 instances) into procurement models.

## Tools
optuna, anthropic.
Optuna optimizes the mathematical allocation between on-demand, reserved, and spot capacity to minimize financial cost.
Hosted a currently available hosted Claude model accessed via API; models and costs are evaluated analytically.

## Deliverables
Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

## Acceptance
- Computes optimal instance blend minimizing overall financial spend for a given SLA threshold.
- Integrates hosted Claude analysis to explain contract commitment risks.
- Outputs clear tabular cost comparisons between procurement scenarios.

## Starter Examples
```json
[
  {
    "input": {
      "workload_type": "ProductionServing",
      "peak_gpus_needed": 32,
      "baseline_gpus_needed": 16,
      "on_demand_rate_hourly": 3,
      "reserved_1yr_rate_hourly": 1.8
    },
    "expected_output": {
      "recommended_reserved_count": 16,
      "recommended_on_demand_count": 16,
      "annual_projected_cost": 294336,
      "procurement_rationale": "Covering the steady 16 GPU baseline with 1-year reserved instances yields a 40% cost reduction on constant load."
    }
  },
  {
    "input": {
      "workload_type": "AdHocResearch",
      "peak_gpus_needed": 8,
      "baseline_gpus_needed": 0,
      "on_demand_rate_hourly": 3,
      "reserved_1yr_rate_hourly": 1.8
    },
    "expected_output": {
      "recommended_reserved_count": 0,
      "recommended_on_demand_count": 8,
      "annual_projected_cost": 52560,
      "procurement_rationale": "Zero baseline usage indicates high variance. Pure on-demand or spot procurement prevents unutilized reserved capital waste."
    }
  }
]
```

## Safety
Procurement simulations are purely financial modeling exercises without issuing cloud reservation API purchases.
Use mock external writes. Require human review before any real deployment.