HS

Harsh Soni

Data Scientist blending AI, Cloud, and ML Ops to solve complex business problems.

About Me

Data Scientist with 5+ years of experience designing scalable machine learning solutions, predictive analytics models, and AI-driven pipelines across various industries. Proficient in Python, R, SQL, and cloud platforms (Azure, AWS) with deep expertise in time series forecasting, NLP (BERT, LLMs), and deep learning (CNN, RNN, GANs).

Proven success in deploying ML models using TensorFlow, PyTorch, and scikit-learn, and integrating them into cloud-native environments via Azure ML Studio and SageMaker. Adept in building explainable models (SHAP), automating ML pipelines with MLflow & Airflow, and translating complex data into business impact through BI tools like Tableau and Power BI.

Core Competencies

Languages

  • Python (NumPy, Pandas)
  • R (ggplot2)
  • SQL
  • SAS
  • Scala

ML & AI

  • TensorFlow, PyTorch
  • NLP (BERT, LLMs)
  • Gen AI, Forecasting
  • Explainable AI (SHAP)

Data Visualization

  • Tableau
  • Power BI
  • Seaborn & Plotly
  • Advanced Excel

Cloud Platforms

  • Azure (ML Studio)
  • AWS (SageMaker)

DevOps & MLOps

  • Git, CI/CD
  • Docker, Kubernetes
  • MLflow, Airflow

Web Technologies

  • Flask, Django, FastAPI
  • REST APIs, JSON

Career Journey

Data Scientist, AI

AIG, United States | Jan 2024 – Present

  • Built and deployed a fraud detection model using LightGBM and PyTorch on over 25 million insurance claims, improving detection precision by 24% and reducing false positives by 19%.
  • Engineered and productionized NLP pipelines using BERT and spaCy for auto-categorization of claim narratives, achieving 93% classification accuracy.
  • Developed forecasting models (ARIMA) for premium income and customer churn, improving financial planning accuracy by 30%.
  • Automated ML workflows using Azure ML Studio, Data Factory, and MLflow, reducing development time by 40%.
  • Designed explainable ML solutions using SHAP and LIME to support model compliance for risk-sensitive portfolios.

Data Scientist, Machine Learning

Adons Softech, India | Jan 2019 – Jul 2022

  • Collected, cleaned, and analyzed data using SQL and Pandas, contributing to a churn prediction model with 87% accuracy.
  • Developed ETL pipelines using Python and Power Query, reducing data preparation time by 60%.
  • Built interactive Tableau dashboards for real-time monitoring of key KPIs.
  • Applied logistic regression and decision tree models, contributing to a 12% lift in quarterly revenue.

Education

Master of Science in Data Science

The University of Texas at Arlington | 2022 – 2024

Bachelor of Technology in Information Technology

Gujarat Technological University | 2015 – 2019

Get In Touch

I'm always open to discussing new projects, creative ideas, or opportunities to be part of an amazing team. Feel free to reach out to me directly or use the contact form.

harshsoni@workmailit.com

+1 (210) 201-5688

Dallas, Texas, United States