Who is Akash Kachhadiya?

Akash Kachhadiya is an AI/ML Engineer at Calyza Tech (Ahmedabad, India) with a B.Com background and a transition into Data Science and AI. He holds three AWS certifications.

What are Akash's AWS certifications?

AWS Certified Machine Learning Engineer – Associate, AWS Certified Solutions Architect – Associate, and AWS Certified AI Practitioner.

What does Akash specialize in?

Machine Learning, Generative AI, Data Science, and building/deploying ML solutions on AWS using Python, SQL, and data platforms like Databricks.

Where is Akash based?

Home town Surat, Gujarat; currently based in Ahmedabad, Gujarat, India.

How can I contact or hire Akash?

Via the contact form on this site, email (akashkachhadiya55@gmail.com), LinkedIn, or GitHub.

Workforce & CompensationTexas State Government Salary Records

Texas State Workforce Compensation & Salary Forecasting

Public sector compensation regression model achieving 0.99999 R² with rigorous outlier isolation

Top Model & Metric

Bagging Regressor & Gradient Boosting

Variance Explained (R²): 0.99999 (99.999%)

5-Fold Cross-Validation R²: 0.99984

Dataset Specs

State agency payroll records

Job class, department, tenure, location

Target: Annual Salary ($ USD)

Validated Outcome

0.99999 R² Score & Sub-Dollar RMSE across validated agency tiers

Verified Notebook Results

Explore Code on GitHub Discuss This Project

Technologies & Machine Learning Frameworks

PythonScikit-LearnBagging RegressorGradient BoostingRandom ForestPandasMatplotlib

01Business & Clinical Problem Context

Public state agencies require transparent, equitable, and competitive compensation structures. Predicting annual salary across hundreds of job titles and multiple state agencies is critical for fiscal budgeting and pay equity audits.

02Data Pipeline & Model Engineering

Built an enterprise regression pipeline. Cleaned PII data, resolved high cardinality across agency and job class titles, performed Interquartile Range (IQR) outlier filtering on non-standard compensation, and evaluated 8 regression models.

03Algorithm Evaluation & Empirical Benchmarks

Algorithm / Model	Train Score	Test Accuracy / Score	5-Fold CV	Status
Bagging Regressor (Best Robust)	0.99995	0.99999 R²	0.99965	best
Gradient Boosting Regressor	0.99999	0.99999 R²	0.99984	best
Decision Tree Regressor	0.99984	0.99999 R²	0.99965	contender
Random Forest Regressor	0.99836	0.99859 R²	0.99672	contender
XGBoost Regressor	0.99615	0.99524 R²	0.99389	contender
AdaBoost Regressor	0.98587	0.98578 R²	0.98512	baseline

Model Selection & Generalization Strategy

Linear regression yielded a theoretical 100% R² due to rigid state pay bands, but this was susceptible to high leverage outliers and non-linear tenure structures. Bagging Regressor (Test R²: 0.99999, Train R²: 0.99995) and Gradient Boosting (Test R²: 0.99999) demonstrated superior robustness on complex salary distributions.

04Quantifiable Business & Operational Impact

Automates compensation equity analysis across state agencies, pinpointing budget discrepancies and ensuring competitive, fair salary offers within 0.01% error bounds.

05Challenges Encountered & Engineering Solutions

1) High cardinality in agency and class titles solved by structured categorical encoding. 2) Outlier compensation values ($11k - $16k) for part-time/contract roles properly identified and handled via IQR bounds to prevent budget model distortion.

06Key Data & Diagnostic Insights

Job classification code and agency budget band account for the majority of compensation variance.

Bivariate analysis showed significant compensation differences by role hierarchy and educational classification.

Isolated part-time outliers from full-time state employee compensation structures.

Available for Machine Learning & Data Science Opportunities

Interested in deploying ML solutions like this?

Whether you are hiring for an ML/AI role or need high-performance predictive pipelines built for your business, let's connect and discuss your requirements.

Contact Akash All Projects

Cardiovascular Heart Disease Risk Diagnostic Pipeline