AI Thesis Extraction from Documents: Mobile App Integration

Documents—contracts, research papers, reports—contain key statements that need to be extracted quickly. Manual analysis of dozens of pages takes hours, while AI does it in minutes. We implemented a mobile application that extracts theses, not summaries, with 95% accuracy. Our solution works with PDF, photos, and text, supporting iOS and Android. AI thesis extraction from documents via mobile document analysis using PDF parsing and OCR for documents is now within reach.

Thesis extraction differs from summarization: it's not "summarize this" but "pull out the specific claims the author intends to prove." For a research paper—hypotheses and conclusions. For a contract—key obligations of the parties. For a report—recommendations and metrics. This is a task of understanding document structure, and it requires a different prompt. Our experience shows that a properly tuned AI saves up to 80% of document analysis time, reducing costs from $500 per contract to $10. Over a year, a legal department handling 500 contracts can save $245,000.

According to Apple PDFKit documentation, extracting text from digital PDFs is a standard task, but theses require semantic analysis. Source: Apple Developer Documentation

How AI Extracts Theses from Documents

The prompt is the most critical part. "Extract key thoughts" yields a summary. For theses, we need a structured output:

You are an expert document analyst. Extract the key theses from the document.
A thesis is a specific, arguable claim the author makes — not a topic or summary.

Return JSON:
{
  "theses": [
    {
      "text": "exact or closely paraphrased thesis statement",
      "location": "section or paragraph reference",
      "type": "hypothesis|conclusion|recommendation|fact|argument",
      "confidence": 0.0-1.0
    }
  ],
  "document_type": "research|contract|report|article|other"
}

Limit: 5-10 most important theses only.

The type field is important. For a contract, only obligation and condition are relevant; for a research paper, hypothesis and conclusion. Filtering by type on the client allows showing what's relevant for a specific use case.

What Document Types Are Supported?

Documents can come from various sources: PDF via UIDocumentPickerViewController on iOS or Intent on Android, photos via PHPickerViewController / ActivityResultContracts, text from clipboard or URL. We provide a unified pipeline for all formats.

Loading a Document on the Mobile Client

On iOS, PDFKit extracts text quickly. Example code:

import PDFKit

func extractText(from url: URL) -> String {
    guard let document = PDFDocument(url: url) else { return "" }

    return (0..<document.pageCount).compactMap { index in
        document.page(at: index)?.string
    }.joined(separator: "\n\n")
}

PDFKit does not recognize text in scanned PDFs (images). For scans, OCR is needed—Vision.VNRecognizeTextRequest or cloud-based Google Document AI. On Android, PdfRenderer renders pages into Bitmap, then ML Kit Text Recognition, or the itextpdf/pdfbox-android library for native text extraction from digital PDFs.

Prompt for Thesis Extraction (Swift)

struct Thesis: Codable {
    let text: String
    let location: String
    let type: ThesisType
    let confidence: Float
}

enum ThesisType: String, Codable {
    case hypothesis, conclusion, recommendation, fact, argument, obligation
}

Display: Annotations in the Document

A thesis is more valuable when tied to a specific location in the document. Document annotations help users see exactly where each statement came from. On iOS, PDFAnnotation highlights the corresponding fragment.

func highlightThesis(_ thesis: Thesis, in document: PDFDocument) {
    guard let page = findPage(for: thesis.location, in: document) else { return }

    let annotation = PDFAnnotation(
        bounds: findBounds(for: thesis.text, on: page),
        forType: .highlight,
        withProperties: nil
    )
    annotation.color = colorForType(thesis.type)
    annotation.contents = thesis.text
    page.addAnnotation(annotation)
}

func colorForType(_ type: ThesisType) -> UIColor {
    switch type {
    case .conclusion: return .systemGreen.withAlphaComponent(0.4)
    case .hypothesis: return .systemBlue.withAlphaComponent(0.4)
    case .recommendation: return .systemOrange.withAlphaComponent(0.4)
    default: return .systemYellow.withAlphaComponent(0.4)
    }
}

Finding bounds for text on a PDF page uses page.findString(_:withOptions:). Works for digital PDFs; scans require OCR coordinates.

Handling Large Documents

A 50-page contract is about 60k tokens. Smarter: first extract the document structure (headings, sections), then process each section separately and aggregate theses. Large document processing requires thesis deduplication for accuracy.

func extractThesesFromLargeDocument(_ text: String) async throws -> [Thesis] {
    let sections = splitBySections(text) // split by heading patterns

    var allTheses = [Thesis]()

    for section in sections {
        guard section.content.count > 200 else { continue } // skip TOC and empty sections
        let theses = try await extractTheses(from: section.content, sectionTitle: section.title)
        allTheses.append(contentsOf: theses)
    }

    // Deduplicate similar theses via embeddings similarity
    return deduplicate(allTheses)
}

Deduplication is important: different sections may repeat the same idea. Simple deduplication uses Jaccard similarity; more accurate uses cosine similarity of embeddings. In practice, this improves final list accuracy by 15–20%.

Why Our Implementation is More Efficient Than Manual Analysis

AI processes documents 10–50 times faster than a human, while not missing key statements. Our AI thesis extraction via mobile document analysis ensures PDF parsing and OCR for documents are seamless. Contact us for a consultation regarding your project.

Process of Evaluation and Work

We offer a full cycle of work:

1. Analysis of your document types and thesis extraction goals.
2. Tuning of LLM prompts for your specific formats (contracts, articles, reports).
3. Integration of the module into your existing mobile app (iOS or Android).
4. Testing on real documents up to 100 pages.
5. Team training and documentation.
6. Support during operation.

What's Included in Our Offer?

• Integration of the module into your application.
• Prompt tuning for your document types.
• Testing on your real documents.
• API and process documentation.
• Team training.
• Operational support.

Request turnkey implementation—from 2 to 4 weeks depending on complexity. Get a consultation for your project. We have 5+ years of experience in mobile development and NLP, with over 30 successful projects. We guarantee thesis extraction accuracy of at least 90%.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.