Senior ML Engineer · Ads & Marketing

Engineering ML systems that
scale and deliver

Machine Learning Engineer building production systems that run the world’s fastest-growing platforms at the relentless pace of modern ad-tech.

Exemplary performance award · 1x Patent
Get in Touch View Work
Machine Learning
Kaggler
Data Engineer
Ads & Bidding Systems
Product Data Scientist
Software Engineer
Shlok Sethia
SHLOK SETHIA
Senior ML Engineer · Walmart · Sunnyvale, CA
Full-stack MLE AI Builder
6+ years
Work experience
$300M+
Revenue Increment
40×
Platform Scale
Top 2%
Kaggle Competition
Career

Work
Experience

As a Machine Learning Engineer I build production end-to-end systems and scalable infrastructure that drive business growth and innovation.

Highlights
Proven Track Record
🏗️
Platform Founding Engineer
One of the first ML engineers on Walmart's vendor-funded advertising platform — scaled it 40× to $300M+ revenue.
⭐
Exemplary Performance Award
Samsung Research recognition for outstanding impact on EfficientNet-B7 computer vision work.
🚀
Promoted in 12 Months
Promoted to Senior Data Scientist at Samsung Research within 12 months — well ahead of cycle.
🛡️
Patent Filed
Novel microarchitecture security work that led to filing a patent in side-channel defenses.
💡
Advanced Software Competency
top-2% performer award for optimized, production-grade code quality.
Walmart Global Tech
Senior Machine Learning Engineer — Ads & Marketing
San Francisco, USA
Jul 2022 – Present
  • Cross-Signal Identity Graph Sequential modeling: Architected and deployed a PyTorch-based user embedding framework to unify highly fragmented, multi-channel data (organic traffic vs. external affiliate clicks) into a cohesive, privacy-compliant identity graph used for sequential modeling, personalized ad targeting and frequency capping.
  • Affiliates Item Recommendation: Led development of an item recommendation tree-based model for Walmart’s $11.3B affiliate market, using a Two-Tower NLP model to generate item name embeddings combined with multi-channel traffic data and competitor metrics, driving higher sales and affiliate partner commissions.
  • Coldstart BERT Model: Built an NLP model using ad metadata to predict conversion rates and RPC (Return per Click) for “cold” items in paid channels, enabling bid optimization, traffic expansion, and identification of high-potential products. Driving 30% increase in traffic and 22% expanded catalog while maintaining ROAS (Revenue Over AdSpend).
  • Top Spender Bidder: Owned the full lifecycle of a bidding system for high spending Product Listing Ads (PLAs) - from design and modeling to algorithm development, A/B testing, deployment, and maintenance - leveraging near real-time data to optimize ad ranking and targeting, control spend, maintain traffic, and improve ROAS.
  • Vendor-Funded Marketing Platform: One of the first ML engineers on the platform — scaling it 40× and driving $300M+ ARR.
Embedding modelsBidding SystemsIdentity GraphsA/B TestingData Engineering
Samsung Research
Senior Data Scientist — Samsung Pay
Bangalore, India
Jun 2019 – Mar 2021
  • Churn Prediction Model: Conceptualized, modeled and deployed an XGBoost classifier to predict users who will churn out in the current quarter. Ran a Grid Search followed by LOFO Importance to attain an accuracy of 83%.
  • Landmark Recognition: Coded an EffNet B7 on PyTorch with Cosine Annealing Scheduler, heavy augmentations, Adams Optimizer, SGD with momentum, and decay to extract Deep Global & Local Features
  • HDBSCAN Clusteringand Heatmap: Screened payment failures on Samsung Pay across big cities in India to identify red zones using HDBSCAN Clustering. Liaised with management and on-field team to ensure zero slippage in critical zones.
  • Time Series Forecasting: Built ARIMA + LGBM ensemble models — 86% accuracy predicting user errors 14 days out based on identity graph patterns
ClusteringCNNPyTorchTensorFlow
Sauma Capital LLC
Quantitative Research Intern
New York City, USA
Jan 2022 – Jul 2022
  • Volatility-Volume Order Slicing: Built an order slicing framework using statistical methods, Markov Chain Monte Carlo (MCMC), and machine learning to optimize execution across market regimes.
  • MCMC Regime Modeling: Applied a Metropolis-Hastings algorithm to estimate transaction probability matrices and approximate regime-dependent distributions.
  • Statistical Trading Research: Developed volatility-volume based execution strategies to improve transaction quality across different market conditions.
Capital MarketsMarkov Chain Monte Carlo (MCMC)Statistical ModelingTime Series
Indian Institute of Science (IISc)
Security Research Scientist
Bangalore, India
Jan 2019 – Jun 2019
  • Novel hardware security research: Investigated side-channel and controlled-channel vulnerabilities in i5-Skylake+ CPUs, advancing novel countermeasure design tied to Meltdown/Spectre-class threats.
  • PAO defense prototype: Developed a first-of-its-kind Page Access Obliviousness implementation in gem5 with TLB-level prefetching to obfuscate page access patterns, driving a patent filing.
  • SGX protection study: Analyzed Intel SGX internals and implemented side-channel and controlled-channel countermeasures.
Hardware SecurityResearch
Academic Background

Education

Strong academic foundation bridging computer science and machine learning research.

Columbia University · New York City, USA
M.S. in Business Analytics
GPA: 3.97 / 4.0 · July 2022
Coursework: Machine Learning, Optimization Models, Probability & Statistics, Marketing Analytics, Algorithmic Trading

Course Assistant: Machine Learning · Tools for Analytics (Python)
RV College of Engineering · Bangalore, India
B.E. in Computer Science
GPA: 9.7 / 10.0 · May 2019
Rank: 3 out of 320 students

Scholarship: Merit-based Infineon Scholarship — awarded to top 1% of students
Stack Profile
Technical
Arsenal

End-to-end ML coverage — from feature engineering and model development to bidding systems, NLP, and production deployment.

{ } Languages
Python SQL (Advanced) PySpark CUDA C++
🧠 ML Tools
PyTorch TensorFlow NLP Transformers LLM
⚗️ Experimentation
A/B Testing Causal Inference Incrementality Testing Power Analysis
☁️ Data & Cloud
Google Cloud Platform BigQuery Hive · Presto Snowflake CassandraDB Elasticsearch
⚙️ Infrastructure
Apache Airflow Docker · Podman Kubernetes Git · CI/CD MLflow
🎯 Domain
Ads & Marketing Bidding and Ranking Systems Recommendation Systems Natural Language Processing Ad Auction Systems
Certifications
Deep Learning Specialization
deeplearning.ai · Coursera
Quantitative Modeling Specialization
Wharton · Coursera
Innovation

Patent
Research

Hardware-level security research targeting Intel's x86 microarchitecture — engineering defences against exploits behind a US$200B+ industry crisis.

PAT-001
PATENT PENDING
OBLIVIAN
Page Access Obliviousness Engine · Intel SGX Defence System

Researched and engineered a software solution to harden Intel i5-Skylake (and above) chips against side-channel and controlled-channel attacks — the class of exploits behind the 2017 Meltdown & Spectre catastrophe that rendered every x86 Intel CPU sold before 2017 vulnerable, costing the security industry US$200+ billion.

Built alongside a team of IISc PhD students and a leading professor from the Indian Institute of Science in Bangalore — India’s premier research and innovation institution — the work leveraged Page Access Obliviousness (PAO) with TLB-level prefetching in gem5 to obfuscate page access patterns attackers exploit through Intel SGX enclaves and eliminate the memory footprint side-channel entirely.

Side-Channel Defence Controlled-Channel Attacks Meltdown / Spectre
Competitions & Projects

Kaggle &
Publications

Competition Expert ranked in the top 2% globally — medal-winning solutions in computer vision and medical imaging.

🥈
Data Science Bowl — Silver Medal

Finished top 3% in the world's largest data science competition with an end-to-end deep learning pipeline.

Weighted KappaEnsemble
View on Kaggle
🥉
APTOS Blindness Detection — Bronze Medal

Detected diabetic retinopathy from rural fundus images — speeding up detection by 36%.

Medical ImagingCNNPyTorch
View on Kaggle
🏆
Object Classification — Best Paper Award

Received Best Paper Award for "Object Classification based on Spatial Orientation" - IEEE publication.

Computer VisionSpatial Orientation
Read Paper
Contact

Let's build something
impactful

Focused on machine learning systems that turn data into reliable decisions, scale models into production, and optimize customer-facing product outcomes.

LinkedIn
linkedin.com/in/shlok-sethia
Email
shloksethia7@gmail.com