United States

AI/ML Engineering

Product Design

Technical Product Management

Back-end

Full stack

Data Engineering

Growth Hacking

Bachelors

Chinmay Nerurkar

Principal AI/ML Engineer

Public Speaking

Product strategy

Communication

Problem-solving

Critical-thinking

Leadership

AI/ML

Business intelligence

Conversational AI

Chatbot

Cloud computing

Machine learning

Postman

Python

React

Airflow

Apache Spark

Axure

Canva

ChatGPT

Docker

GitHub

Java

Jupyter Notebook

gRPC

REST API

Redis

PostgreSQL

Kafka

•	Building cutting-edge Gen AI capabilities for Microsoft partners.

AI2030 Global Fellows is a prestigious and exclusive program designed to bring together the brightest minds from around the world who are passionate about the future of Artificial Intelligence (AI) and its impact on society. The fellowship aims to create a diverse and collaborative community of experts, researchers, and practitioners who will collectively work towards shaping the development and responsible use of AI over the next decade.

•	Leading a multitude of strategic engineering initiatives across the Microsoft Advertising Org on Audience & Identity, Ad Attribution & Measurement through privacy compliant, data-driven, AI solutions at scale. Xandr conducts over 400 Billion daily real-time Ad auctions world-wide with an average response time below 250 milliseconds.
•	End-to-end owner of planning, engineering architecture, design and delivery of Backend systems, UI/APIs, Monitoring, Introspection & Product Release.
•	Leading Engineering on enhancing Ads Ranking to improve RoAS and First-Party Audience Modeling to extend advertiser reach by 10-20x using Graph Neural Network (GNN) & LLM (GPT) models.  
•	Leading efforts to mitigate impact of third-party cookie deprecation (3PCD) using contextual modeling for cookie-less targeting, measurement and attribution.
•	Collaborating on company-wide Targeting Data Reduction effort to remove low-value targeting segments from the auction process workflow for 50%+ reduction in compute & network utilization across the ad-serving stack for significant cost savings.
•	Providing technical & thought leadership to address Ad-campaign budget underspend and drive more revenue on Xandr Invest DSP by novel solutions to improve targeting, supply forecasting, detecting incorrect campaign setup, etc.

•	Built Advanced Frequency Management product, an innovative frequency-capping feature that uses site/app co-visitation probabilities for unique users and auction win-rate estimates for Ad campaigns to let Advertisers control the number of impressions seen by non-addressable (“cookie less”) users.
•	Led cross-company efforts with Xandr and WarnerMedia to build an Identity Graph for Cross-Device Targeting with 11 Billion Nodes & 16 Billion Edges using a proprietary Householding technique based on HashMin and modified Gale-Shapley algorithm. 
•	Led company-wide efforts in making Digital Ad-business compliant with French Data Protection Agency (CNIL) interpretation of GDPR. Working closely with Privacy and Legal teams to keep Xandr compliant with CCPA/CPRA, VCDPA and newer privacy laws.
•	Managed cloud migration efforts to move Audience and Identity team’s on-prem tech-stack to MS Azure cloud, with a focus on Azure DB performance and low-latency SLAs for Xandr’s real-time applications.
•	Oversaw company-wide hiring efforts to fill mid-level, senior roles for engineering, data science and product management teams. Mentoring engineers across the organization, helping them set and achieve career growth goals.

•	Led design and implementation of Xandr Invest DSP’s Buy-side ML Product Suite for click & conversion prediction helping Advertisers to achieve their Ad-campaign KPIs across digital, mobile, video and CTV inventory. 
•	Authored and open-sourced ElasticNet4j, a performant Logistic Regression ML library written in Java, customized to handle sparse input data and extremely rare outcomes like clicks & conversions.
•	Designed and built K8s-Workqueue, a pluggable scheduling mechanism for ML workloads on Kubernetes & Docker in collaboration with a multi-disciplinary engineering team and data scientists. K8s-Workqueue has accelerated product velocity while reducing development complexity and currently manages 200+ ML/ETL/Analytics workloads.
•	Presented K8s-Workqueue at PAPIs ML conference in 2019, Identity Resolution using Graphs at Graph+AI conference in 2020. Participated on industry panels to discuss Machine Learning Engineering and Graph technology at Xandr. 
•	Managed cloud migration efforts to move Audience and Identity team’s on-prem tech-stack to MS Azure cloud, with a focus on Azure DB performance and low-latency SLAs for Xandr’s real-time applications.
•	Recognized as a Tech Community leader for taking ownership of Xandr’s tech blog, building an enablement team to facilitate creation of blog posts by Xandr employees, spanning a wide spectrum of technical & non-technical topics.

•	Streamlined Petabyte scale Dataflows by building unified ETL pipelines to serve Data Science R&D and Xandr ML products.
•	Provided design and architectural oversight on critical engineering projects across multiple functional capabilities (streaming, big-data, ML pipelines, etc.) to successfully improve software quality, system reliability and stability.
•	Achieved 50% reduction in resource utilization and over $2 Million in saved costs by leading companywide efforts to reconfigure and optimize operation of database (Vertica & TigerGraph) and big-data ecosystem (Hadoop MR, Spark, etc.)
•	Created Java/Scala framework for complex event processing of 100 Billion+ streaming events per day with metrics, monitoring, Kafka & DB connectivity out of the box. Widely used by multiple teams at Xandr with 15+ applications running in production. 
•	Built Ad Spend-Pacing system using Lambda architecture to manage Ad campaign dollars spent across 1000+ bidders in 5 datacenters in the Americas, Europe and Asia with a latency of <20 seconds, eliminating Millions of dollars in yearly overspend.

•	Designed and built Market-data and Low Latency Order Management System for US equity trading in Java. The OMS handled over 3% of all US equities traded daily by volume.
•	Worked on developing a revolutionary Algo Trading Engine employing smart execution tactics built around acute knowledge of market micro-structure to achieve over 20% improvement in execution quality. 
•	Built RPC/messaging infrastructure over TCP (ZeroMQ) and reliable UDP Multicast (with TCP retransmission for only-once, in-order messaging) using Google Protocol Buffers in Java & C++.
•	Led a team on an industry-first, trade surveillance product to catch & prevent trading fraud using streaming order data analysis to discern market-manipulative patterns, providing feedback to the OMS to restrict/block 'bad' traders, all in real-time.
•	Built a process to analyze daily trading patterns and process metrics in trailing windows to dynamically set/tweak risk, exposure and edge control parameters per trader for each day

•	Worked on the Streaming Event Processing team, built software infrastructure processing Terabytes of financial data daily for bringing transparency to the fixed-income markets by providing real-time bond pricing with high accuracy.
•	Developed messaging libraries for reliable Publish/Subscribe over varying network conditions with client API in C++ and Scala/Java using ZeroMQ and Google Protocol Buffers.

•	Worked on a team to build high-definition IP camera products using multi-threaded C/C++ in Linux on ARM processors.
•	Designed an intuitive web UI for responsive Pan-Tilt-Zoom (PTZ) camera movement control while streaming H.264 video over IP with real-time motion detection. 
•	Managed the product development, engineering and integration effort with sub-contractor teams working across 3 continents.

Mentorship Session

Chinmay Nerurkar🇺🇸

About Chinmay Nerurkar

Expertise

Sessions with Chinmay Nerurkar