Skip to content

03_LLD.md

Latest commit

 

History

History
186 lines (152 loc) · 5.4 KB

03_LLD.md

File metadata and controls

186 lines (152 loc) · 5.4 KB
Raw

🧩 Low-Level Design (LLD) 🧩

Context

Project Overview

Field Details
Project Name US Visa Approval Prediction
Prepared By Md. Alahi Almin Tansen
Revision Number 1.0
Last Revision Date 02-09-2025

⬆️ Go to Context

Version History

Revision Date Changes Made Author
1.0 02-09-2025 Initial Draft Md. Alahi Almin Tansen

⬆️ Go to Context

1. Project Structure

US Visa Approval Prediction/
├── components/      # ML pipeline modules: data ingestion, transformation, validation, trainer, pusher
├── config/          # YAML / JSON configuration files
├── configuration/   # Python config loader
├── constants/       # Global constants
├── data/            # raw / processed / interim datasets
├── entity/          # entities like artifact & config objects
├── logger/          # logging setup
├── notebooks/       # EDA / experimentation
├── pipeline/        # training & prediction pipelines
├── tests/           # unit & integration tests
├── utils/           # helper functions
├── app.py           # demo or API entry point
├── demo.py
├── requirements.txt
├── Dockerfile
└── setup.py

⬆️ Go to Context

2. Key Modules & Responsibilities

Module Responsibility
components/ ML workflow: ingestion → validation → transformation → training → evaluation → deployment
pipeline/ Orchestrates training & prediction steps
services/ High-level APIs: PredictService, TrainService
configuration/ Reads environment, config files, paths
logger/ Central logging
data/ Stores raw/interim/processed data

⬆️ Go to Context

3. Visualization: Class Diagram

classDiagram
    class Config
    class Preprocessor
    class ModelWrapper
    class Registry
    class PredictService
    class TrainService

    Config <.. Preprocessor
    Config <.. ModelWrapper
    Registry o-- ModelWrapper
    Registry o-- Preprocessor
    PredictService --> Registry
    TrainService --> Registry
Loading

⬆️ Go to Context

4. Data Layer (MongoDB)

  • Collections

    • applications: stores raw applications with features & target (case_status)
    • models: stores model metadata, version, artifact paths, metrics

  • ERD

    erDiagram
APPLICATIONS {
    int id PK
    varchar continent
    varchar education
    varchar job_experience
    int num_employees
    float prevailing_wage
    int age_of_company
    int applicant_age
    varchar case_status
}
MODELS {
    int id PK
    varchar version
    varchar model_path
    varchar preproc_path
    json metrics
}
    Loading

⬆️ Go to Context

5. Orchestration Sequences

  • Prediction Flow

    sequenceDiagram
    participant C as Client
    participant API as FastAPI
    participant S as PredictService
    participant R as Registry
    participant M as Model
    participant P as Preprocessor

    C->>API: POST /predict {features}
    API->>S: validate + predict()
    S->>R: get_latest()
    S->>P: transform(df)
    S->>M: predict_proba(X)
    S-->>API: {decision, probability, version}
    API-->>C: 200 JSON
    Loading

  • Training Flow

    sequenceDiagram
    participant OPS as CI/Runner
    participant API as FastAPI
    participant T as TrainService
    participant R as Registry

    OPS->>API: POST /train {data_uri, algo}
    API->>T: start pipeline
    T->>T: ingest -> split -> preprocess.fit -> model.train
    T->>R: register(model, preproc, metrics)
    T-->>API: {version, metrics, paths}
    Loading

⬆️ Go to Context

6. Deployment Topology

flowchart LR
    Client -->|HTTPS| Ingress[API Gateway/Ingress]
    Ingress --> API[FastAPI Service]
    API --> Reg[Model Registry]
    API --> Logs[(Log Store)]
    API --> Metrics[(Prometheus)]
    Reg --> Vol[(Artifacts Volume/S3)]
Loading

⬆️ Go to Context