AI Safety & Security
Senior Software Engineer — AI Safety, Security & ML Infrastructure
Securing AI systems at scale. I build systems that make generative AI safer and more secure, grounded in deep production ML infrastructure experience. My work spans AI security detections and telemetry, LLM evaluation, model inference, and high-throughput data platforms.
01
Safety first, proven at scale
AI Safety & Security
Lead a compliant AI security platform that turns product signals into safety and security detections and enriched telemetry across Microsoft 365 Copilot, Azure AI Foundry, and Microsoft AI.
Detection & response
Served as a key engineering driver during a high-severity AI security incident, rapidly delivering detection dashboards and indicator pipelines while partnering across security teams to support investigation and containment decisions.
LLM evaluation
Co-led an LLM evaluation and benchmarking framework with applied scientists and engineers, adopted across Bedrock Agents for launches including Multi-Agent and Code Interpreter.
ML infrastructure
Led four engineers in building a Dockerized Python inference sidecar and Java service, migrating Alexa Query Rewriting to native AWS, enabling autoscaling, and reducing host costs by approximately $3M annually.
Conversational AI
Led model-based contextual query rewriting and slot carryover that helped Alexa preserve context across conversational turns, understand follow-up requests without customers repeating information, and deliver more natural interactions across nine locales. Screen-based context resolution improved continuity between what customers saw and asked.
Prime Video operations
Drove cross-team ML data classification and compliance automation that raised operational stability across Prime Video businesses. A supporting Spark and Python pipeline processed billions of records daily while saving $50K annually.
02
Security focus, infrastructure depth
Lead and subject-matter expert for a multi-region, multi-tenant AI Security Platform spanning Microsoft 365 Copilot, Azure AI Foundry, and Microsoft AI. Build compliant ML data pipelines, security detections, and telemetry systems; partner across organizations on incident response; and mentor engineers delivering AI security and compliance capabilities.
Lead engineer and primary point of contact across generative AI, ML inference, and contextual conversation teams. Delivered LLM evaluation, model-serving, contextual AI, and large-scale data systems while leading cross-team designs and mentoring engineers.
03
Capabilities
Responsible AI, prompt-injection and memory-poisoning detection, indicator detection, security telemetry, investigation tooling, incident response, Kusto and KQL, and Grafana.
LLMs and agents, PyTorch, AllenNLP, model inference serving, runtime batch inference, evaluation and benchmarking, autoscaling, and latency and cost optimization.
Azure ML, Databricks, Spark and PySpark, Azure Data Factory, AWS, Docker, Python, Java, SQL, distributed systems, and high-throughput pipelines.
Design ownership, cross-team and cross-organization collaboration, mentoring, engineering excellence, interviewing, hiring, monitoring, alerting, and dashboarding.
04
Academic foundation
2015 — 2017
Data Science & Intelligent Systems
University of Texas at Dallas
2009 — 2013
SASTRA University, India
Let’s connect
Connect with me to discuss AI Safety & Security, production ML infrastructure, distributed systems, and technical leadership.