CV

ML Systems Engineer

Google Cloud

Inference optimization and high-performance serving of Gemini models across GPUs and TPUs.

• Driving inference optimization and serving performance for Gemini models across GPU and TPU platforms.
• Previously led the TPU v7 (Ironwood) NPI microbenchmarking suite; bridged internal cluster infrastructure (Borg) with Cloud XLA runtimes, ported benchmarks from HLO to JAX, and executed multi-cube scaling analyses to model accelerator interconnect limits.
• Delivered collective communication microbenchmarks and public recipes for the TPU v7 public preview, establishing interconnect scaling baselines that advanced multi-billion-dollar hardware evaluations with tier-1 frontier AI customers.
• Spearheaded intra-node disaggregated serving for Gemini models on bare-metal nodes in Google Distributed Cloud (GDC) / Sovereign Cloud.
• Architected automated continuous regression pipelines ensuring Borg-Cloud runtime parity; optimized open-source LLMs (Llama 3, DeepSeek) to land Day-1 3P hardware enablement (Sim-Ship), earning the Google Cloud Tech Impact Award (2026).

Applied Scientist

AWS, Amazon

Architected Generative AI-powered observability and diagnostics for big data systems and network security at hyperscale.

• Architected and launched a Generative AI-powered auto-debugging system (Amazon Q) for Apache Spark across AWS Glue, Amazon EMR, and MaxDome, utilizing LLMs for automated root-cause analysis to slash Mean Time to Resolution (MTTR) from days to minutes.
• Engineered a high-throughput active testing service for AWS network security, injecting synthetic live traffic to validate Network ACLs at scale, guaranteeing strict compliance and eliminating false-positive blindspots inherent to static simulation models.
• Designed and integrated automated diagnostic tooling directly into AWS Glue and EMR management consoles, delivering real-time failure insights and actionable remediations to thousands of enterprise customers.

Research Intern

Microsoft Research, Redmond

Networking Research Group.

• Architected low-latency distributed protocols for edge-cloud deep learning inference offloading within the Networking Research Group, minimizing communication overhead and end-to-end latency under bandwidth-constrained networks.

Research Fellow

Microsoft Research, India

Systems and networking research for edge computing, IoT, and AI.

• Co-designed and built FarmBeats, an end-to-end IoT and cloud analytics platform for data-driven agriculture; awarded Microsoft Outstanding Technical Achievement Award and productized into Microsoft Azure (US Patent US20200150640A).
• Pioneered DeepEdge, a novel edge-AI scheduling architecture optimizing deep learning workloads under severe compute and bandwidth constraints, enabling graceful performance degradation (US Patent US10942767B2; published at NSDI).
• Researched neural network learnability with theoretical researchers, analyzing empirical bounds on how data complexity, model depth, and minibatch dynamics govern generalization.
• Mentored 4+ research interns and junior engineers, driving collaborative systems research initiatives that resulted in multiple conference publications.

Ph.D.

Carnegie Mellon University

Electrical & Computer Engineering

• Thesis: Practical Network Layer Machine Learning for IoT Security
• Advisors: Prof. Vyas Sekar & Prof. Anthony Rowe

B.Tech and M.S. by Research

International Institute of Information Technology (IIIT)

Computer Science

• Computer Vision
• Broadcast Video Analytics

Google Cloud Tech Impact Award

Google Cloud

Awarded for landing Day-1 3P hardware enablement (Sim-Ship) and optimizing open-source LLMs (Llama 3, DeepSeek) with runtime parity.

Microsoft Outstanding Technical Achievement Award

Microsoft

Awarded for co-designing and building FarmBeats, productized into Microsoft Azure.

CIT Dean's Fellow

Carnegie Mellon University

Doctoral fellowship awarded by the College of Engineering at Carnegie Mellon University.

Dean's List for Academic and Research Excellence

IIIT Hyderabad

Awarded for academic and research excellence during B.Tech & M.S. by Research.

Lumen: A Framework for Developing and Evaluating ML-Based IoT Network Anomaly Detection

ACM CoNEXT

Designed an ML evaluation framework across 100+ devices, resolving domain-shift degradation in network traffic anomaly detection.

Lumos: Identifying and Localizing Diverse Hidden IoT Devices in an Unfamiliar Environment

USENIX Security

Built a mobile system leveraging 802.11 PHY/MAC features and AR to localize hidden IoT devices with 95% accuracy; covered by Hacker News front page, Der Spiegel, and CyLab.

Contention-Aware Performance Prediction for Virtualized Network Functions

ACM SIGCOMM

Developed an ML-driven contention-aware predictor for virtualized network functions (VNFs) to guarantee QoS under multi-tenant cloud workloads.

All that GLITTERs: Low-Power Spoof-Resilient Light Anchors for Augmented Reality

ACM IPSN

Engineered an optical retro-reflective sensing pipeline enabling sub-centimeter AR localization at milliwatt power consumption.

DeepEdge: A Network Edge for Deep Learning Workloads

USENIX NSDI

Designed a distributed edge-cloud scheduling framework optimizing DNN inference pipelines under tight compute constraints.

Deep neural network workload scheduling

US Patent 10,942,767

Scheduling framework for optimizing deep learning inference workloads across heterogeneous edge and cloud compute under tight resource constraints.

Sensor fall-curve identification

US Patent 11,327,476 (US20200150640A1)

Time-of-flight and RF signal fall-curve analytics for agricultural sensing and IoT infrastructure (Microsoft FarmBeats).

Lumos: Identifying and Localizing Diverse Hidden IoT Devices in an Unfamiliar Environment

USENIX Security • Boston, MA

Lumen: ML-Based IoT Network Anomaly Detection

CyLab & CONIX Annual Review • CMU, Pittsburgh, PA

All that GLITTERs: Light Anchors for Augmented Reality

ACM IPSN / CPS-IoT Week • Sydney, Australia (Virtual)

FarmBeats: AI, Edge & IoT for Agriculture

Microsoft Techfest • Redmond, WA

Automated Top-View Registration of Broadcast Sports Video

IEEE WACV • Lake Tahoe, NV