{
  "docId": "019dd923-605d-726d-81a4-062653b9a1ac",
  "docSlug": "bi-35336c8b45abb80f",
  "documentTitle": "Check out the decks 4 AI startups used to raise millions from VCs",
  "authorId": "Pitchdecks",
  "authorName": "Qevlar AI",
  "documentKindSlug": "pitchdeck",
  "documentKindLabel": "Pitch deck",
  "sourceTypeSlug": "strategy_consulting",
  "sourceTypeLabel": "Strategy consulting",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.5,
  "pageNumber": 22,
  "pageCount": 42,
  "prevPage": 21,
  "nextPage": 23,
  "slideType": "solution",
  "function": "present_solution",
  "density": "overcrowded",
  "nDataPoints": 0,
  "notes": "The slide uses a three-part process flow to explain the product's evaluation capabilities.",
  "elementsJson": [
    "headline_text",
    "action_title",
    "paragraph",
    "bullet_list",
    "screenshot"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-605d-726d-81a4-062653b9a1ac/22",
  "deckHref": "/decks/019dd923-605d-726d-81a4-062653b9a1ac",
  "deckJsonHref": "/decks/019dd923-605d-726d-81a4-062653b9a1ac.json",
  "deckAnchorHref": "/decks/019dd923-605d-726d-81a4-062653b9a1ac#slide-22",
  "components": [
    {
      "bbox": {
        "h": 0.2,
        "w": 0.35,
        "x": 0.53,
        "y": 0.53
      },
      "kind": "image",
      "text": null,
      "attrs": null,
      "subkind": "screenshot",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "229e1fe6-c0c7-4677-8ae8-7fdc565375a9",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.35,
        "x": 0.53,
        "y": 0.75
      },
      "kind": "image",
      "text": null,
      "attrs": null,
      "subkind": "screenshot",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "76cd1618-86a2-46f0-b5ad-1ac693d2a53a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.35,
        "x": 0.53,
        "y": 0.28
      },
      "kind": "image",
      "text": null,
      "attrs": null,
      "subkind": "screenshot",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "c6972818-10fa-4de1-b619-8b2a8503d20b",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.4,
        "x": 0.11,
        "y": 0.75
      },
      "kind": "list",
      "text": "#3: Human Review\nCollect human feedback from subject matter experts\n- Custom evaluation criteria with editable fields directly in the UI\n- Flexible field types: binary, numeric, and string options\n- Keyboard shortcuts ← → to quickly navigate across datapoints",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "17379ead-882b-4b40-8aeb-8449039860cd",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.25,
        "w": 0.4,
        "x": 0.11,
        "y": 0.28
      },
      "kind": "list",
      "text": "#1: Setup Evaluators\nBuild custom metrics in UI / code\nDefine evaluators (i.e. metrics) to test and detect failure modes:\n- Code Evaluators: Create simple assertions and unit tests in Python or use external packages (Transformers, Scikit Learn, etc.)\n- LLM Evaluators: Evaluate semantic qualities like Toxicity, Coherence, Answer Faithfulness, etc. using other LLMs + your own policy/criteria\n- Human Evaluators: Create custom annotation fields and criteria for domain experts to review",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "97ecc66d-0cb0-4398-bb26-a96e589ba0d8",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.2,
        "w": 0.4,
        "x": 0.11,
        "y": 0.53
      },
      "kind": "list",
      "text": "#2: Run Experiments\nRun experiments to evaluate quality\n- Log experiments programmatically via our SDK or in CI/CD using GitHub Actions/Jenkins/CircleCI\n- Compare results from any two experiments over the same test suite\n- Analyze evaluation report, slice and dice results, and share learnings with your team",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f3ca68db-a160-4700-9577-1d3a098f3226",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.11,
        "y": 0.22
      },
      "kind": "paragraph",
      "text": "Test and benchmark performance of your application on a component-level every time you make changes. Run ad-hoc experiments with your team or integrate into your existing CI/CD workflows.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "56baa2d9-5ae3-45a6-b926-ab676fa628d3",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.5,
        "x": 0.11,
        "y": 0.11
      },
      "kind": "title",
      "text": "Run automated evals to ship with confidence",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "4d3cc62f-0f70-495d-a334-e51f5c2881c7",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "List presentation",
      "slug": "list-presentation",
      "agent": null,
      "layer": "slide",
      "matchId": "6fa41887-b9eb-42ad-ac03-90b5c0c31dc9",
      "evidence": "The slide contains a list/bullet with multiple items.",
      "confidence": 0.9
    }
  ],
  "frameworks": [],
  "arcBeats": [],
  "loops": [],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}