I will build machine learning models with explainable predictions

India

I speak Tamil, English

Full Stack Web and AI Agent Developer

Hi, I'm Saran, a full stack and AI developer. I build fast, clean web apps with React, Next.js and FastAPI, AI agents and RAG chatbots that automate real work, and Python scripts that save you hours. ...
About this Gig

Need predictions you can actually trust and explain to your team?


I build machine learning models for classification, scoring and forecasting: churn, fraud, failure, lead scoring, risk and more. Every model comes with honest metrics and clear explanations of WHY it predicts what it does.


What you get:

  • Data cleaning, feature engineering and leak-free validation
  • - Models with scikit-learn, XGBoost, LightGBM or PyTorch, tuned for your goal
  • - Honest metrics for imbalanced data (precision, recall, F1, AUC-PR), not just "accuracy"
  • - SHAP explanations showing which features drive each prediction
  • - A plain-English report, plus a prediction API and dashboard in Premium

Proof:

In the SanDisk hackathon my team built WaferLens, predicting which chips will fail with 7.9x better AUC-PR than the classical baseline, plus per-chip SHAP explanations. My Song Genie engine identifies songs from 80,000+ tracks using Bayesian inference. Both are in my portfolio.


Send me a message with your dataset and goal before ordering, and I'll tell you honestly what's achievable. Let's build it!

Expertise:

Feature learning

•

Classification

•

Churn

•

Decision trees

Programming language:

Python

•

SQL

•

Colab

Frameworks:

Scikit-learn

•

Keras

•

PyTorch

Tools:

Jupyter Notebook

•

OpenCV

•

TensorFlow

My Portfolio