# EdgeTal Complete Platform Documentation & Specifications
Developed by Knivok Private Limited (DBA Knovik) · https://knovik.com

===============================================================================
1. PLATFORM OVERVIEW & ARCHITECTURE
===============================================================================

EdgeTal turns your smartphone, tablet, or laptop into a 100% self-contained AI recruiter. Candidate resumes, search queries, notes, and AI evaluations remain strictly inside your device hardware sandbox.

Key Specs:
- Legal Publisher: Knivok Private Limited (DBA Knovik), Colombo, Sri Lanka.
- Zero Cloud Transmission: No third-party LLM APIs, no analytics telemetry, no remote databases.
- 512d MediaPipe Embeddings: Dense text vector representations generated locally.
- Gemma-2B IT GenAI Engine: Quantized Int4 model running on Apple Metal (MPS) and Android GPU delegates.
- Hardware Platform Targets: iOS (Build 1.0.1+6 TestFlight Public Beta), macOS (Apple Silicon & Intel), Android.
- Data Security: AppLockService biometric Face ID / Touch ID vault lock, password-encrypted .edgetal archives.

===============================================================================
2. STEP-BY-STEP RECRUITER WORKFLOW
===============================================================================

Step 1: Import Candidates in Seconds
- Local Folder Scan: Select local folders containing PDF or DOCX files.
- Cloud Drives: Direct file picker integration for iCloud Drive, Google Drive, and OneDrive.
- CSV Datasets: Load structured candidate rows via CSV file or web URL.
- Under the hood: EdgeTal extracts text and generates 512-dimensional vector embeddings on-device.

Step 2: Create Open Job Roles & Competencies
- Define job title, location, required skill tags (e.g. Flutter, Kafka, PostgreSQL), and job description.
- EdgeTal builds a 512d job vector representation in local ObjectBox database for real-time candidate ranking.

Step 3: Run Natural Language Talent Searches
- Type natural search phrases ("Senior Full-Stack Engineer with microservices and high-scale Postgres").
- Sub-100ms vector matching ranks candidates by semantic similarity percentage with exact keyword highlighting.

Step 4: Execute On-Device Gemma-2B AI Fit Analysis
- Select job role and run on-device LLM analysis.
- Outputs: Step-by-step reasoning thought process, verifiable CV evidence quotes, and definitive shortlist/maybe/pass verdict.

Step 5: Manage Pipeline & Share Encrypted Packages
- Drag candidates across pipeline stages: Shortlisted -> Interviewing -> Offer -> Placed.
- Export password-encrypted .edgetal packages for peer-to-peer sharing over AirDrop, USB, or email.

===============================================================================
3. NATIVE HARDWARE ACCELERATION ENGINE
===============================================================================

iOS & macOS Architecture:
- Embedding Engine: Swift MediaPipeTasksText CocoaPod creating 512d dense vector representations.
- LLM Acceleration: Metal Performance Shaders binding directly to GPU & Apple Neural Engine.
- Security: Protected with iOS Keychain and AppLockService Face ID / Touch ID vault locks.

Android Native Architecture:
- Embedding Engine: Native Kotlin EmbedderChannel.kt with Google TFLite text embedder bindings.
- LLM Acceleration: LlmChannel.kt over com.google.mediapipe:tasks-genai (GPU & NNAPI delegates).
- Local Database: High-performance ObjectBox HNSW C++ vector engine storing local candidate embeddings.

===============================================================================
4. APPLE TESTFLIGHT PUBLIC BETA (BUILD 1.0.1+6)
===============================================================================

iOS Public Beta is live on Apple TestFlight!
How to join:
1. Provide your iCloud Email address (or Apple ID) on edgetal.com.
2. Receive your official Apple TestFlight invite email.
3. Accept invitation in the TestFlight app on iPhone, iPad, or Mac.
4. Install EdgeTal Build 1.0.1+6 and experience 100% on-device AI recruiting.

Requirements:
- iPhone: iOS 14.0+, 3GB RAM recommended.
- iPad: iPadOS 14.0+ with Apple Neural Engine or Metal GPU support.
- Mac: macOS 11.0+ (Big Sur) on Apple Silicon or Intel processors.

===============================================================================
5. RESEARCH & EMPIRICAL VALIDATION
===============================================================================

EdgeTal's architecture is published in a peer-reviewed research paper presented at EICON 2026 (ESOFT International Conference). The system was empirically evaluated across two hardware tiers with benchmarks on retrieval accuracy, latency (<90ms), scalability (2,484+ resumes tested on-device), and generative analysis quality.