About me

Puneet Singh Ludu

Machine Learning Architecture & Innovation

Building Scalable AI Systems for Enterprise Impact

View Detailed Resume

Professional Overview

At TRDDC (Tata Research Development and Design Centre), I pioneered algorithms for event detection in time series data, developing a Shape Context-based solution that achieved 7% improvement over traditional methods like SAX and DTW. This work established my foundation in practical machine learning applications.

My early contributions included architecting a Data Harmonization Framework (DHF) utilizing map-reduce for real-time enterprise data integration. These experiences shaped my approach to machine learning architecture: building robust, scalable systems that deliver measurable business value.

Leadership & Technical Innovation

ML Architecture at Zillow

Leading the Zestimate team's machine learning architecture, where I spearheaded the development of an interactive Comparative Market Analysis (CMA) platform. This initiative leverages Siamese Neural Networks to provide data-driven valuation tools, establishing new revenue channels while enhancing user engagement.

Orchestrated the modernization of our ML infrastructure through Python, Terraform, AWS, Kubeflow, and Metaflow, achieving $500k annual cost optimization and 95% reduction in operational incidents.

Revenue Optimization at OkCupid

Engineered subscription pricing optimization systems utilizing Wide & Deep learning architectures. Implemented comprehensive ML pipelines with Python, Keras, and TensorFlow, resulting in 6% revenue growth through systematic A/B testing and dynamic pricing strategies.

Enterprise Solutions at FactSet

Developed mission-critical systems including a CNN-based speaker identification platform for earnings calls, an ELMo/BiLSTM-powered company information extraction system, and an optimized document processing service achieving 66% performance improvement. Enhanced FactSet terminal's formula discovery, improving average rank precision from 5.6 to 2.3 through advanced language modeling.

Technical Approach & Methodology

My approach to machine learning systems is founded on three core principles:

Performance-Driven Innovation

Focusing on quantifiable outcomes: from infrastructure cost optimization ($500k annual savings) to revenue enhancement (6% growth through ML-driven pricing).

Architectural Excellence

Designing systems for scale and reliability, demonstrated through successful implementations in high-throughput environments at Zillow, OkCupid, and FactSet.

Technical Leadership

Contributing to the ML community through conference organization (MUFin Workshop at AAAI2023, PKDD2022) and mentorship initiatives.

Key Achievements

Enterprise ML Systems

Architected and deployed an interactive CMA platform with real-time valuation capabilities, driving significant revenue growth through enhanced agent tooling and market analysis.

ML Infrastructure

Implemented advanced speaker identification systems and document processing pipelines, achieving 20% operational efficiency improvement and 66% reduction in processing time.

Open Source Development

Developed high-impact tools including Lotion (2K+ GitHub stars), Romadeva (adopted by Translators Without Borders), and jTextBrew for advanced string matching.

Research & Development

Academic Publications

  • Latent Attribute Inference in Social Networks (ACM Hypertext 2015)
  • Gender Classification via Celebrity Network Analysis (CORR 2014)
  • Automated Music Mood Classification System (IJCSI, 2010)

Technical Projects

  • MUFin Workshop - Financial Uncertainty Modeling (AAAI2023, PKDD2022)
  • Resume Analysis System - NLP-based Career Development Tool
  • Quena - Large-scale Question Answering System (1.6M Document Index)

Academic Background

M.S. Computer Science

State University of New York at Buffalo (2014)

B.Tech. Computer Science

Jaypee Institute of Information Technology (2010)

Professional Network

Open to discussions on machine learning architecture, system design, and innovative technical solutions.