Sumit Mamtani

Sumit Mamtani’s career sits at a demanding junction of modern computing: identity, security, and large-scale platform reliability—areas where small engineering improvements can translate into outsized business impact and meaningful user benefit. Since 2015, Mamtani has built a profile that combines high-scale industry delivery with research contribution and professional service. Across roles in global consumer platforms and advanced academic training, he has consistently focused on translating computer science into measurable outcomes—improving authentication, fraud prevention, performance, and user experience for millions of people—while also contributing to the broader technical community through publications, peer review, and mentorship.

At Walmart Global Tech (2022–present), Mamtani works within the Identity organization, responsible for authentication, fraud prevention, and customer experience at massive scale. His work reflects a consistent emphasis on conversion, security, and operational excellence. He led the implementation of phone and email OTP during account creation and contributed to passkeys (FIDO2/WebAuthn), helping improve sign-in conversion by approximately 3.5%. He also helped deliver high-impact sign-in experience improvements—such as a one-tap password-manager flow and phone-based OTP—each credited with roughly +0.25% site-wide GMV impact, quantified at approximately $132M per year per feature.

Beyond product outcomes, Mamtani’s work includes platform efficiency and reliability—an area often invisible to users but central to sustainable scale. He led JVM warm-up and tuning work (REME) in collaboration with Azul, reducing pod requirements from roughly 90 to 40, improving reliability while saving an estimated $245K annually. He also contributed to modernization initiatives such as migration to WCP-Profile, retirement of legacy APIs, and removal of redundant calls—preparing the identity stack for multi-tenant growth without compounding technical debt. More recently, he designed an extensible KYC platform using Cosmos DB and modular vendor providers (including Experian and SheerID), targeting major reductions in vendor dependence while improving throughput, accuracy, and latency. In parallel, he has been building an agentic AI end-to-end test runner using synthetic identities to execute cross-tenant login and consent flows—an effort aimed at strengthening identity assurance and regression safety in complex, multi-tenant systems.

These contributions earned formal recognition: in October 2024, Mamtani received Walmart’s Engineering Excellence (Bravo) award from Walmart Global CTO Suresh Kumar—an endorsement that reflects both technical impact and organizational trust.

Before Walmart, Mamtani’s tenure at Swiggy (2019–2021) marked a transition from ML research to high-scale backend engineering under real traffic constraints. He led the design and development of a widget-based homepage capable of handling roughly 100K concurrent requests with P99 latency around 20ms. He engineered core services sized for approximately 1M RPM, migrated eventing to Kafka, and delivered schema and caching changes that reduced on-call burden by about 45% while saving up to $100K per year. He also shipped a Spring Boot OCR-based “food intelligence” service used by approximately 100K restaurants—an example of applied ML integrated into production workflows. This period shaped his operational instincts: balancing latency, reliability, and cost while maintaining development velocity.

Mamtani strengthened his research depth through graduate studies at NYU Courant (M.S., 2021–2022), specializing in machine learning, NLP, and deep learning. During a 2022 internship at Walmart Global Tech, he re-architected identity services on Apollo GraphQL/Node and improved P99 latency by approximately 20% at ~220K concurrent users, while adding full observability and production readiness for token validation—work that foreshadowed his later identity platform contributions.

Alongside industry delivery, Mamtani has maintained a research and publication track, treating research and production as a reinforcing loop. He has published peer-reviewed work in venues including BMVC (2018) and reports additional publications in 2025 spanning transformer-based vision optimization and graph-integrated text classification. He also notes ongoing submissions and works under review, including topics such as neural weather nowcasting, communication-efficient split federated learning for LLMs, and contrastive pretraining for fine-grained classification. This dual footing—platform engineering and active research engagement—positions him to bring emerging methods into practical systems, while also grounding research direction in real operational constraints.

His community service record is substantial and sustained. Mamtani regularly reviews for international venues (including IEEE IJCNN and other conferences in computer vision, pattern recognition, and NLP) and serves on a Reviewer/Advisory Committee for the Soft Computing Research Society (SCRS), contributing to program curation and peer review. He maintains active membership in professional bodies including IEEE, ACM, INNS, and BCS—demonstrating ongoing commitment to professional standards and cross-community engagement.

A parallel thread in his profile is mentorship and team leadership. Mamtani mentors interns and junior engineers and trains peers in high-impact operational skills—P99 latency reasoning, code quality practices, DevOps fundamentals, ML/DL foundations, JVM warm-up behavior, and SLO design. He also leads cross-functional execution across product, security, and frontend stakeholders, emphasizing design documentation, fair code reviews, and high-quality debugging practices as mechanisms to raise capability across teams, not merely deliver one-off features.

The earliest evidence of his “paper-to-production” habit appears during his B.Tech at IIT Hyderabad (2015–2019). In 2017, he built and deployed a helmet-detection system from traffic video as a Django REST API reportedly used by a city, achieving approximately 95% accuracy and demonstrating a public-safety application of applied ML. In 2018, he interned at Swiggy as an ML researcher and improved a CNN for food classification by about 30% through architectural enhancements. These early projects framed a career pattern that continues today: turning research and engineering into systems that scale, serve users, and withstand operational reality.

Across identity, security, platform efficiency, applied ML, and research service, Mamtani’s work reflects a Fellow-caliber posture: high-scale outcomes, defensible technical leadership, and consistent contributions to the broader community.