Dr. Maya Krishnan

Data Scientist & Researcher

I turn data into actionable insights.

I'm a data scientist and ML engineer with a PhD in Statistics and 6 years of industry experience. I specialize in building production ML systems that drive real business outcomes. My work spans NLP, computer vision, and causal inference, with publications at NeurIPS and ICML.

I built a model to optimize my coffee brewing ratio

My Jupyter notebooks are embarrassingly well-organized

I once spent a week debugging a model only to find a data labeling error

I dream in confusion matrices

Senior Machine Learning Scientist

RetainIQ

Mar 2021 — Present · San Francisco, CA

Lead the customer intelligence ML team building churn, propensity, and lifetime-value models. Shipped a gradient-boosted churn predictor (94% accuracy) that saved $2.3M in annual retention costs, and established the MLOps practice covering feature stores, model monitoring, and automated retraining.

Python XGBoost MLOps FastAPI AWS SageMaker

Data Scientist

LogiChain Supply Co.

Jul 2018 — Feb 2021 · Austin, TX

Built demand-forecasting and anomaly-detection systems across a global supply chain. Delivered a real-time analytics dashboard tracking 15 KPIs and a time-series forecasting model that cut inventory holding costs by 18% across three regional warehouses.

Python Prophet Apache Kafka Tableau SQL

Research Scientist Intern

AI Research Lab

May 2017 — Aug 2017 · Remote

Researched fine-tuning strategies for transformer-based language models. Co-authored a workshop paper on transfer learning for low-resource sentiment classification and prototyped the multilingual NLP pipeline later productionized in industry.

PyTorch Hugging Face NLP Research

Programming

Python / Pandas 95%

R 85%

SQL 90%

Jupyter / Notebooks 92%

ML & AI

TensorFlow / PyTorch 85%

Scikit-learn 90%

NLP 78%

Computer Vision 72%

Visualization

Matplotlib / Seaborn 90%

Tableau 80%

D3.js 70%

Customer Churn Prediction Model

Gradient-boosted model predicting customer churn 30 days in advance with 94% accuracy. Deployed via real-time API serving 100K predictions/day. Saved $2.3M in annual retention costs.

Python XGBoost scikit-learn FastAPI

Sentiment Analysis NLP Pipeline

Fine-tuned BERT model for multi-label sentiment analysis across 12 languages. Processes 500K social media posts daily with an F1 score of 0.91.

PyTorch Hugging Face SpaCy Docker

Real-time Analytics Dashboard

Interactive data visualization dashboard for supply chain metrics. Real-time streaming data, anomaly detection alerts, and drill-down capabilities across 15 KPIs.

D3.js Python Apache Kafka Tableau

Causal Inference in Ad Spend — Research Paper

Published research on causal inference methods for digital advertising attribution. Accepted at NeurIPS 2024 workshop. Novel approach to measuring incremental lift without A/B tests.

R Stan Bayesian Statistics LaTeX

35+ Models in Production

12 Research Papers

50+ Datasets Analyzed

94% Best Model Accuracy

PhD — Statistics

Stanford University

2013 — 2018

Dissertation on Bayesian methods for causal inference in observational data. Published two first-author papers at NeurIPS and ICML and served as a teaching assistant for graduate-level machine learning.

BS — Mathematics & Computer Science

University of Michigan

2009 — 2013

Graduated summa cum laude with a double major. Conducted undergraduate research in applied statistics and led the campus data science club.

Causal Inference for Ad Attribution

NeurIPS 2024 Workshop (Causal ML) 2024

Transfer Learning for Low-Resource Sentiment Classification

ICML 2019 Proceedings 2019

AWS Certified Machine Learning — Specialty

Amazon Web Services 2022

Let's Build Something Data-Driven

Have a dataset that needs unlocking or a model that needs shipping? I collaborate on research, ML consulting, and production data science. Reach out and let us talk about your problem.

Download vCard

Create Your Digital Business Card

Free — no signup required 0 contact cards created

Fill in your details to generate a downloadable contact card (vCard) and QR code. Works on all phones. No data is stored.

Personal Info

Full Name *

Work

Job Title / Role

Company / Organization

Contact

Email

Phone

Website

Social Profiles

Twitter / X

GitHub

Instagram

Address (optional)

Short bio / note

Private by design: your vCard and QR code are generated entirely in your browser. None of the details you type are ever transferred to, used by, or stored on this website. The only thing recorded is an anonymous +1 to the total number of cards created — never any personal data.

Dr. Maya Krishnan

Data Scientist & Researcher

About

Experience

Skills