Praviveek Ray
Praviveek Ray · Data Scientist

Turning complex data into confident decisions.

Data Scientist with 1.5+ years of experience at Mahindra Comviva, delivering machine learning, advanced analytics and AI solutions at enterprise scale. My work spans the full lifecycle — from raw data to deployed models — with outcomes measured in millions of users and real revenue.

Start a project View my work Download resume ↓
Find me on LinkedIn GitHub
Praviveek Ray
Measured impact
0M+
telecom users impacted by production ML models
+0%
renewal conversion through AI-driven segmentation
−0%
post-to-prepaid migration via targeted frameworks
0%
accuracy in published deep-learning research (GCN)
Experience

Professional experience.

A steady progression at Mahindra Comviva — from AI/ML training to owning production data science for one of India's largest telecom platforms.

Oct 2024 — Present

Mahindra Comviva

Data Scientist
Bengaluru, India · Full-time
  • Reduced subscriber attrition by 6% and lifted recharge transactions 5% across a 70M+ user base through production machine learning models.
  • Improved renewal conversion by 22% and cut post-to-prepaid migration 16% by designing AI-driven segmentation frameworks for 60M+ customers.
  • Delivered analytics solutions across 4+ cross-functional projects, streamlining ETL pipelines that power personalized marketing at scale.
  • Built interactive Power BI dashboards that translate model outputs into clear insights for business stakeholders.
  • Developed an agentic AI solution that autonomously cleans, analyzes and extracts key attributes from complex datasets.
PythonSnowflakePower BIAgentic AIETL
Jan 2024 — Jul 2024

Mahindra Comviva

Data Scientist Intern
Bengaluru, India · Internship
  • Built and deployed a multimodal AI chatbot handling text, images and PDFs, improving user engagement through NLP and computer vision.
  • Engineered deep learning solutions with transformers and CNNs, deployed on Google Cloud for scalability.
  • Automated customer-support workflows by integrating external systems and APIs, validated through rigorous benchmarking.
FlaskTransformersCNNsGoogle Cloud
Jul — Aug 2023

Mahindra Comviva

AI/ML Trainee
Bhubaneswar, Odisha · Training
  • Predicted customer churn at 81% accuracy for a telecom client using multiple ML models on a 200K-row dataset with SQL, Python and advanced Excel.
  • Reduced manual reporting effort by ~1.5 hours per cycle through automated analysis workflows.
SQLPythonAdvanced Excel
Featured projects

Selected projects.

Industry and independent work across machine learning, deep learning, GenAI and analytics — each with source code and documentation.

Graph structure · sentiment nodes
Deep Learning · NLP · Published research

IMDb Sentiment Analysis with GCN

Published research combining a Graph Convolutional Network with Word2Vec and TF-IDF embeddings for movie-review sentiment, achieving 88.32% accuracy with a robust preprocessing pipeline.

Code → Research paper →
Multimodal pipeline
Generative AI · Multimodal

Gemini Chatbot — Multimodal Assistant

A production chatbot that understands text, images and PDFs, combining transformer-based NLP with CNN image processing — deployed on Google Cloud and integrated with external systems.

● Live demo → Code → Report →
Churn probability · declining trend
Machine Learning · Classification

Telecom Customer Churn Prediction

End-to-end churn modeling on a 200K-row telecom dataset — multiple ML models benchmarked to enable targeted retention offers, with automated reporting workflows.

Project →
Conversion funnel · retail analytics
Computer Vision · Analytics Platform

Store Intelligence System

CCTV-to-conversion retail intelligence: five camera feeds fused with POS data to produce footfall counts, a monotonic conversion funnel, dwell analytics and explainable anomalies — served via FastAPI with a live dashboard.

● Live demo → Code → Pitch deck →
Model comparison · ANN vs RNN vs LSTM
Deep Learning · Benchmarking

Bank Churn Prediction with Deep Learning

Customer churn prediction for banking, benchmarking three deep learning architectures — ANN, standard RNN and LSTM — to compare how each captures patterns in customer behavior.

Code →
Behavioural clusters · 6 segments
Unsupervised ML · Analytics

Customer Segmentation Model

Six behavioural segments recovered from an unlabelled subscriber base with K-Means and a Gaussian Mixture Model, a supervised churn classifier layered on top, and an interactive dashboard for tuning the decision threshold.

● Live dashboard → Code →
Dual-engine RAG · retrieval trace
GenAI · Retrieval-Augmented Generation

Glasshouse — A RAG Pipeline You Can See Through

Two answer engines over one shared retrieval trace: an extractive engine that quotes sources verbatim and cannot hallucinate, beside a generative engine that writes — so the difference is attributable to answering alone.

● Live demo → Code →
More on GitHub
Weathero — weather app Pravi Library — book manager DSA problem solutions
View all repositories
Expertise

Core competencies.

The languages, methods and platforms behind the results above.

Languages

  • Python
  • SQL
  • R
  • Java
  • JavaScript
  • HTML / CSS

ML & AI

  • Machine Learning
  • Deep Learning
  • NLP & Transformers
  • Generative & Agentic AI
  • Computer Vision
  • A/B Testing & Statistics

Data & Cloud

  • Snowflake
  • Google Cloud
  • AWS
  • Hadoop · Hue
  • PostgreSQL · MySQL
  • ETL Pipelines

Visualization

  • Power BI
  • Tableau
  • Advanced Excel
Licenses & certifications

Verified credentials.

Every certificate below links to its official issuer record or credential document for verification.

Simplilearn
Issued Dec 2023

Introduction to Cloud Computing

Credential ID4715327
Status✓ Verifiable
View certificate
Udemy
Issued Oct 2023

C# and .NET Core with Real-Life Examples

Credential IDUC-5666f9e5…f42a0
Status✓ Verifiable
View certificate
IBM · Coursera
Issued Oct 2023

IBM Authorized Course on Cyber Security

Verified byCredly badge
Status✓ Verifiable
View badge
Cisco Networking Academy
Issued Sep 2023

Cisco Certified Cyber Security Course

Verified byCredly badge
Status✓ Verifiable
View badge
Intellipaat
Issued Sep 2023

Certified DevOps Course

Credential ID31679-1655-137586
Status✓ Verifiable
View certificate
GeeksforGeeks
Issued Aug 2023

Google Cloud Bootcamp

FocusGCP · Compute · IAM
Status✓ Verifiable
View certificate
BSNL
Issued Jul 2023

Industrial Internship on Basic IT

Credential ID2023/BIT/R12/128/02
Status✓ Verifiable
View certificate
Now writing

Signal.

A weekday AI & tech digest, arXiv paper picks, and my own research notes — filtering the signal out of the noise.

Read it ↗
Freelance engagements

Have a data problem? Let's solve it together.

I take on select freelance engagements — translating your requirements into working, well-documented solutions with clear milestones.

  1. Share your requirementsEmail me the business goal, the data you have (or don't), and your timeline.
  2. Receive a scoped proposalA clear plan within days: approach, deliverables, milestones and effort estimate.
  3. Iterative build & reviewDelivery in checkpoints — models, dashboards or pipelines, with documentation you can maintain.
  4. Handover & supportClean handoff of code, reports and a walkthrough, plus follow-up support.

What I can build for you

Predictive models · customer analytics · GenAI & chatbot systems · computer vision solutions · ETL pipelines · business intelligence dashboards · experiment design & analysis.

Fixed scopeMilestone basedRemote friendly
Send your requirements
Contact

Recruiting, collaborating, or just curious?

I reply fastest on email. All channels below are active.

Emailpraviveek@gmail.com Phone+91 82604 60214 LinkedIn/in/praviveek-ray GitHub@21pravi