# Pavan Rangudu > SDE II (Senior) @ AWS Elemental | MLOps • GenAI Infrastructure • Distributed Systems | CMU | IIT KGP Location: Sunnyvale, California, United States Profile: https://flows.cv/pavanrangudu I’m an SDE II at AWS Elemental, working where video infrastructure meets MLOps and AI/ML. I spend my days making machine learning systems production-ready, optimizing inference pipelines, hardening evaluation frameworks, and turning research prototypes into reliable distributed systems. Recently, I’ve been focused on our new AI video analysis platform (launching soon). I re-architected VLM (Vision Language Model) evaluation workflows using SageMaker and Triton Inference Server, cutting latency by 96% and also built fast-fail guardrails that prevent thousands of wasted compute jobs monthly. I work closely with Applied Scientists to productionize open-source vision models, and I contributed to an Innovation Award-winning RAG (Retrieval-Augmented Generation) pipeline for video search using LanceDB vector database and SigLIP multimodal embeddings, enabling semantic retrieval over large media datasets. Before AWS, I spent two years at Amobee building real-time data pipelines for ad targeting via Spring Boot microservices handling thousands of events per second and TB-scale datasets for look-alike audience modeling. I’m strongest where distributed systems meet ML infrastructure. I love the MLOps side making models fast, reliable, and cost-efficient at scale but I’m just as comfortable building high-throughput backend services. Heavy on Java (Spring Boot) and Python, deep in AWS (Lambda, DynamoDB, SageMaker), and always looking for hard problems at the intersection of infrastructure and AI. M.S. Software Engineering from Carnegie Mellon, B.Tech from IIT Kharagpur. ## Work Experience ### Software Dev Engineer II @ AWS Elemental, an Amazon Web Services Company Jan 2022 – Present | Santa Clara, California, United States • Designed and operated RESTful control plane APIs for MediaConvert (high-scale video transcoding service), processing millions of daily job orchestration requests across distributed microservices (AWS API Gateway, Lambda, DynamoDB, SQS, S3, ECS/EC2). • Led cross-team initiative on Step Functions and MediaConvert integration across CreateJob and Probe APIs, driving ~20% user growth; mentored engineers to production readiness. • Delivered high-visibility integrations for NAB 2024 by collaborating with Solutions Architects, Product, and partner teams; recognized for 'Customer Obsession' and 'Invent & Simplify' Leadership Principles during quarterly business reviews. • Designed and launched a secure file-sharing API for customer support investigations, reducing resolution time from days to hours while meeting strict AppSec and compliance requirements. • Built fraud containment, authorization enforcement, and fast-fail validation systems, restricting 500+ high-risk accounts and preventing thousands of failures per month while reducing wasted compute. • Partnered with applied science teams to productionize ML inference pipelines; re-architected SageMaker-hosted model evaluation workflows, cutting inference evaluation latency by ~96%. • Contributed to an Innovation Award-winning project by building vector-based video search using SigLIP embeddings and scalable inference pipelines. ### Lead Software Engineer @ Amobee Jan 2022 – Jan 2022 | Redwood City, California, United States • Developing a Customer Data Platform to onboard, amplify & monetize 1st party data to target look-alike users for advertisements • Led an international team of 3 members to automate the amplification of 1st party data of an online delivery platform over 6 sprints in an Agile development process. This entailed daily standups, presentations to stakeholders, code reviews, and system testing • Collaborated with Engineering, Product Management & Data Operations teams to develop an algorithm using 3rd party Graphs that helped our data pipelines more efficiently process user data and enhance the campaign creation rate by 20% • Leveraged Spark, Kafka & Microservices to ingest real-time advertising data into a low-latency data store for Bid Optimization • Implemented Data Deduplication feature that reduced business intelligence data by 60% • Trained and mentored new interns & full-timers in improving their Spark Scala skills & get familiar with our big data pipelines ### Senior Software Engineer @ Amobee Jan 2020 – Jan 2022 | Redwood City, California, United States ### Graduate Teaching Assistant @ Carnegie Mellon University Jan 2019 – Jan 2019 | Mountain View, California ### Graduate Research Assistant @ Carnegie Mellon University Jan 2019 – Jan 2019 | Mountain View, California Worked on a NASA - CMU collaboration project aimed at developing a web application to provide NASA crucial insights on the advancement of science and research focus areas in the Earth Science domain. As a part of the project, I designed and developed a standalone dockerized Java application that facilitates the training of topic model - Labelled LDA (Latent Dirichlet Allocation) using Stanford Topic Modelling Toolbox (TMT). This involved data extraction, feature engineering, NLP based Data analysis of research publications. Technologies Used: Scala, Java with Play framework, MongoDB ### Applications/Software Engineering Intern @ Intel Corporation Jan 2019 – Jan 2019 | Santa Clara, California, USA Worked on designing and developing software tools and frameworks to improve the performance, throughput, and reporting of Front End Tools, Flows and Methodology. Update/Write automation scripts via Bash scripting, Python, Perl & Tcl. ### Senior Software Development Engineer @ Siemens EDA (Siemens Digital Industries Software) Jan 2016 – Jan 2018 | Fremont, California, United States Developed production quality QuestaSim software using C/C++ as a Kanban team of 10 to accurately model the effects of active power management(UPF) on SV RTL designs; facilitated Semiconductor companies to efficiently verify their designs Proposed and implemented various low-power coverage methodologies in C/C++ in Linux platform; leveraged multithreading, multi-processing concepts; made QuestaSim stand out from competitive simulation tools in terms of functionality & performance DV Conference US18 paper: Took the initiative to analyze and resolved the mystery of low-power designs code coverage ## Education ### Master of Science - MS in Computer Software Engineering Carnegie Mellon University Silicon Valley ### Bachelor of Technology - BTech in Electrical Engineering Indian Institute of Technology, Kharagpur ## Contact & Social - LinkedIn: https://linkedin.com/in/pavanrangudu --- Source: https://flows.cv/pavanrangudu JSON Resume: https://flows.cv/pavanrangudu/resume.json Last updated: 2026-04-11