Hey there! I'm Puneet

Senior Machine Learning Engineer at Zillow's Zestimate team, where I architect and lead the development of AI-powered real estate valuation systems. With over 11 years of experience across fintech, dating, and real estate domains, I specialize in building production-scale ML systems that drive business impact. From optimizing pricing algorithms that generated millions in revenue to developing real-time speaker identification systems, I thrive on transforming complex data challenges into scalable solutions.
New York MS Computer Science Open Source Contributor
Profile Picture

Technical Expertise

Technologies I work with

Languages

Python Java C/C++ Bash JavaScript SQL

ML/AI

PyTorch TensorFlow Keras Scikit-learn Kubeflow Metaflow

Infrastructure

AWS Docker Terraform MongoDB PySpark FastAPI

Career Journey

My professional experience and education

Machine Learning Engineer

Zillow (Zestimate) | 2021 - Present

Leading the architecture and development of Zillow's next-gen valuation systems. Spearheaded the interactive CMA platform development and infrastructure modernization, achieving $500k annual cost savings. Mentoring team members and driving technical innovation across the organization.

Machine Learning Engineer

OkCupid (Match.com) | 2020 - 2021

Led end-to-end development of ML pricing optimization system, implementing sophisticated pricing models that drove 6% revenue increase through A/B testing. Owned the complete ML pipeline from feature engineering to production deployment.

Machine Learning Engineer

FactSet | 2015 - 2020

Architected multiple ML systems including real-time speaker identification for earnings calls, automated company information extraction from 1.6M websites, and near-duplicate document detection service. Reduced compute time by 66% and pioneered new search ranking algorithms.

Master of Science in Computer Science

State University of New York, Buffalo | 2014

Focused on Machine Learning and Natural Language Processing. Published research on user attribute inference and social network analysis.

ML Research Engineer

Tata Research Development and Design Centre | 2011 - 2013

Developed novel algorithms for time series pattern detection achieving 7% accuracy improvement over existing methods. Built an ETL framework for real-time enterprise data harmonization using MapReduce paradigm.

B.Tech in Computer Science and Engineering

JIIT, India | 2010

Foundation in Computer Science and Software Engineering. Published research on automated music mood classification.

Notable Projects

Key projects across professional, academic, and open-source domains

Professional

Interactive CMA Platform

Zillow • 2023

Led development of an ML-powered Comparative Market Analysis platform using Siamese Neural Networks, enabling real-time property valuations and automated comps selection.

Python PyTorch AWS
Dynamic Pricing Engine

OkCupid • 2021

Architected an ML system for subscription pricing optimization using Wide & Deep learning, achieving 6% revenue growth through personalized pricing.

TensorFlow MLOps A/B Testing
Speaker Identification System

FactSet • 2019

Built a real-time speaker identification system for earnings calls using CNNs, processing 100K+ hours of audio with 95% accuracy.

Deep Learning Audio Processing Real-time

Research & Academic

Quena - Question Answering System

SUNY Buffalo • 2014

Developed a large-scale QA system indexing 1.6M documents, using advanced NLP techniques for semantic search and answer extraction.

NLP Information Retrieval Java
Social Network Analysis Framework

ACM Hypertext • 2015

Created a framework for inferring user attributes through celebrity network analysis, published at ACM Hypertext 2015.

Graph Analysis Machine Learning Python
Music Mood Classification

IJCSI • 2010

Pioneered an automated system for music mood classification using audio feature extraction and ML, published in IJCSI.

Signal Processing Classification MATLAB

Open Source

Lotion

2K+ GitHub Stars • 60K+ Downloads

Unofficial Notion.so Desktop app for Linux, bringing the full Notion experience to Linux users with native integration.

Electron JavaScript Linux
Romadeva

Used by Translators Without Borders

Advanced Roman to Devanagari script converter with contextual awareness and linguistic rule processing.

NLP Python Linguistics

Research Publications

My contributions to academic research

Inferring Latent Attributes of an Indian Twitter user using Celebrities and Class Influencers

ACM Hypertext 2015

Architecture for Automated Tagging and Clustering of Song Files According to Mood

IJCSI, 2010