TEKSUN PROCESS

End-to-End Product Development & Project Fulfillment

Teksun's 5 Phase AI Development Life Cycle

Teksun follows a proven Life Cycle for AI projects, encompassing all crucial stages in bringing a product to life, from ideation to market launch. At every step of the process, Teksun provides unwavering support to ensure a successful outcome.

Phase 1

DESIGN

Phase 2

DEVELOP

Phase 3

TUNE

Phase 4

AUTOMATE

Phase 5

MAINTAIN

Key Activities

Defining Acceptance Criteria
Data Source Finalization
Model Architecture Identification
POC If Required

Key Deliverable

Product Requirements Specification
Architecture Document

Key Activities

Data Annotation
Data Pre-Processing
AI Model Training
Model Transformation for the Target Platform
Inference Code for Target Platform
Happy Path Testing

Key Deliverable

ETL Pipeline
AI model v1.0
Model Benchmarking Report for Test Data vs Live Data

Key Activities

Hyperparameter Tuning Of Model
Benchmarking Of Models
Evaluate Secondary Matrics If Required
Risk Assessment
Quantization And Pruning Of Model

Key Deliverable

AI Model V2.0
Model Benchmarking Report For Test Data Vs Live Data

Key Activities

MLOps
Automate Data Collection
Automate Data Labeling
Automate Model Training
Automate Testing and Deployment
Automate Data Drift Identification

Key Deliverable

Cloud Infrastructure
Automated 360 Pipeline
Data Labeling Review Setup

Key Activities

Auto Scaling For Cloud Deployment
Continuous Monitoring Of Model Performance
Drift Correction Through New Dataset Addition
Drift Correction Through Hyper Parameter Tuning

Key Deliverable

Scalable Cloud Infrastructure
Up To Date Model Performance

4-8 WEEKS

12 - 24 WEEKS

4 - 8 WEEKS

ONGOING

Phase 1

DESIGN

Key Activities

Defining Acceptance Criteria
Data Source Finalization
Model Architecture Identification
POC If Required

Key Deliverable

Product Requirements Specification
Architecture Document

4-8 WEEKS

Phase 2

DEVELOP

Key Activities

Data Annotation
Data Pre-Processing
AI Model Training
Model Transformation for the Target Platform
Inference Code for Target Platform
Happy Path Testing

Key Deliverable

ETL Pipeline
AI model v1.0
Model Benchmarking Report for Test Data vs Live Data

12 - 24 WEEKS

Phase 3

TUNE

Key Activities

Hyperparameter Tuning Of Model
Benchmarking Of Models
Evaluate Secondary Matrics If Required
Risk Assessment
Quantization And Pruning Of Model

Key Deliverable

AI Model V2.0
Model Benchmarking Report For Test Data Vs Live Data

4 - 8 WEEKS

Phase 4

AUTOMATE

Key Activities

MLOps
Automate Data Collection
Automate Data Labeling
Automate Model Training
Automate Testing and Deployment
Automate Data Drift Identification

Key Deliverable

Cloud Infrastructure
Automated 360 Pipeline
Data Labeling Review Setup

4 - 8 WEEKS

Phase 5

MAINTAIN

Key Activities

Auto Scaling For Cloud Deployment
Continuous Monitoring Of Model Performance
Drift Correction Through New Dataset Addition
Drift Correction Through Hyper Parameter Tuning

Key Deliverable

Scalable Cloud Infrastructure
Up To Date Model Performance

ONGOING

Data Source (Options)

Commercial Datasets Open Source Datasets Simulate Datasets Through Third Party Licencing Tools Dedicated Team for Data Collection

RFQ to FDR- Teksun Project Delivery Cycle

Below you will discover the various stages of your Project’s Delivery Cycle, starting from a Request for Quotation (RFQ) which initiates
the Product Requirement Gathering process, and ending with a Final Delivery Report, which releases all final Project Deliverables.

01.