Architecture and design lead for an ML-driven optimization system that personalizes message delivery timing at large scale. Authored the system's North Star architecture, redesigning how user-preference signals flow through ranking, pacing, and infrastructure layers -- moving from a constraints-first model to a user-centric model where downstream systems derive from aggregated user preference.

Areas of focus:

ML ranking and optimization at population scale

Cross-team architecture: aligning modeling, infrastructure, calibration, quality, and pacing teams around a unified design

Designing systems that balance short-term business outcomes with long-term user value

Authoring canonical design specifications that anchor multi-quarter team roadmaps

Working across: modeling teams, distribution infrastructure, auction and pacing systems, privacy-preserving optimization for unmatched user populations, and partnership on revenue-vs-experience tradeoff analysis.

Recent work has emphasized translating senior-engineer-led architecture into staged milestone delivery with measurable business outcomes.

Lead technical and organizational strategy for Meta’s Anti-Scraping Privacy area, overseeing strategy and direction of 4 teams (~30 engineers) tackling large-scale, adversarial data scraping problems across Meta’s family of products.

Key Achievements:

• Scaled Anti-Scraping mitigation effectiveness by over 200% through the introduction of adaptive ML-driven defenses, reducing bad actor persistence across Meta surfaces.

• Built and grew an ML engineering team from 4 → 10 within one year, defining technical charters, and onboarding processes for long-term sustainability. Retention stands at near 100% over the last 3 years.

• Drove an org-wide restructuring from functional silos to a problem-focused structure, significantly reducing redundant efforts (support the same amount of work with 30% less people).

• Launched a new Discovery team (8 engineers) to proactively identify emerging scraping behaviors using LLM-based behavioral analysis — scaling label output by 20x while cutting human review cost by 90%.

• Filled leadership gap for departed org tech lead, defining last 2 half’s technical roadmaps and strategy for 6 teams, ensuring continuity and meeting regulatory requirements.

• Partnered with Privacy, Product Growth, and Security teams to define company-wide scraping risk tolerance and budget tradeoffs, driving more sustainable defenses.

Led Gmail’s Inbound Protections team responsible for safeguarding billions of enterprise and consumer users from spam, phishing, and abuse. Directed the redesign of Gmail’s aging classification pipeline to enable faster, more accurate, and more adaptive abuse detection.

Selected Achievements:

• Modernized the 17-year-old Gmail Abuse Classification System, decomposing legacy components into modular services — eliminated years of tech debt, and cut root-cause iteration time from multiple days to minutes.

• Launched an autonomous abuse monitoring framework leveraging pattern detection and anomaly scoring to proactively flag emerging threats, reducing scaled incidents by 80%.

• Influenced industry-wide adoption of anti-abuse email standards (DMARC, BIMI, ARC) — contributed to spec design and ecosystem rollout, directly enhancing sender authenticity and reducing spoofing attempts globally.

• Mentored and up-leveled team of 6 engineers, establishing a growth-oriented culture that improved engagement and retention scores year-over-year.

Build the vision and team for the research function at McAfee’s Data & Insights Group (previously McAfee Labs), driving research & development of McAfee’s future products, projects & programs.

Projects:

• Drive the effort for improving detection effectiveness of McAfee’s machine learning based anti-malware engine expanding coverage to new threats and greater predictive prowess.

• Design & develop the next-generation analytics platform utilizing multi-enterprise product telemetry to identify and detect emerging threat patterns for timely protection.

Led the industry-leading research at McAfee for signature-less identification of malicious programs using Machine Learning which resulted in substantial improvement in zero-day detection efficacy of McAfee’s products, also allowing reduction in disk footprint leading to customer delight and retention. This is a key product component called Advanced Threat Protection which since 2017 is a highlighting part of McAfee’s both enterprise and consumer products. My contributions to the solution:

• Method for Featurization of Behavioral Events of Programs.

• Method for Incremental Clustering for Identification of Malware Variants in real-time.

• Method for Signature-less Classification of Programs using Regression & Support Vector Machines.

• Method for Profiling of Malicious Processes using Decision Trees (USPatent# 9323928)

• Method for Identifying Malicious Programs based on Icon Similarity in Portable Executables.

Led design and development of multiple high-impact endpoint & cloud projects at McAfee.

Real Protect: Novel and industry-leading cloud-based Anti-Malware Solution for classification of objects using behavior-based and static features. Success of this system led to increased revenue and McAfee’s step into cloud-based products. Major contributions included:

• Key Architect & Developer for scalable backend cloud in AWS for providing real-time classification to clients. Designed using the micro-service architecture allowing cost-effective independent scaling of components capable of handling throughputs of over 10,000 queries per second.

• Collaborating with multiple-geo cross-functional teams including executives and product managers ensuring quality and timely delivery of the product.

• Developer for classification & featurization modules in the endpoint client.