Aug 2023 – Nov 2024 · 1 yr 3 mos · Los Angeles, California, United States · Remote

• Architected and delivered two mission-critical systems that transformed the company's MLS data platform, directly contributing to 30% revenue growth (exceeding $100M ARR) by eliminating the CEO-identified top pain point.
• Real-Time MLS Data Streaming Platform:
• Designed and implemented an end-to-end Kafka-based streaming architecture that reduced listing sync times from no SLA to <15 minutes for 400+ MLS providers with diverse APIs, formats, and protocols. Built a fully automated pipeline using Airflow for 5-minute API polling, Spark Structured Streaming for transformation, and DynamoDB for state management. Optimized Confluent Kafka costs through strategic partition management and implemented comprehensive monitoring for throughput and system health. Deployed on Kubernetes with CircleCI CI/CD, handling deduplication, backfills, schema evolution, and data standardization across petabyte-scale data lakes.
• AI-Powered MLS Triage System:
• Developed a multi-agent AI platform using PydanticAI and LLM frameworks to automate data discrepancy investigations that previously consumed hundreds of manual hours weekly. Built natural language query capabilities over internal databases (Athena, PostgreSQL) and external sources (Zillow, Redfin) with RAG-based context retrieval using LanceDB vector store. Implemented SQL generation with multi-gate validation pipeline, integrated Slack bot interface with DynamoDB conversation persistence, and created specialized agents for listing comparisons, field mapping explanations, and automated bug categorization.
• Technologies: Python, AWS (MWAA/Airflow, EMR, EKS, Athena, DynamoDB, S3), Kafka/Confluent, Kubernetes, Docker, Spark, CircleCI, OpenAI/LLMs, PydanticAI, Protobuf, LanceDB, FastAPI, Starrocks

May 2023 – Aug 2023 · 3 mos · Cupertino, California, United States · Remote

• AI/ML Platform Standardization & Kubernetes Migration - Team 1
• Performed organization-wide migration of Apple's AI/ML data platform to standardized Kubernetes architecture on AWS EKS. Migrated distributed computing stack including Apache Flink, Spark, Kafka, Hadoop, Jupyter, Trino, and Ranger to containerized infrastructure. Rewrote data pipelines for cloud-native execution with Iceberg/Delta Lake on S3, optimized for Parquet storage patterns.
• Key Contributions:
• Designed EKS cluster architecture with VPC networking, NLB ingress, and IAM-based access controls
• Converted legacy Hadoop MapReduce and Flink streaming jobs to Kubernetes-native deployments with stateful checkpoint management
• Built Prometheus/Grafana observability stack with custom dashboards tracking job SLAs, resource utilization, and data freshness
• Implemented automated remediation workflows (Python/Shell) triggered by alert conditions
• Created CI/CD pipelines and self-service tooling for data scientists to deploy independently
• Identified cost optimizations reducing AWS spend by 30%+ through rightsizing and reserved capacity
• Collaborated with 10+ AI/ML teams ensuring zero-downtime migration. Created documentation and training enabling teams to self-manage Kubernetes workloads post-transition.
• Tech: Kubernetes (EKS), Flink, Spark Streaming, Kafka, Hadoop, Iceberg, Delta Lake, Trino, Jupyter, Ranger, AWS (EC2, S3, VPC), Prometheus, Grafana, Python, Helm
• Airflow Platform Modernization (Ads Infrastructure) - Team 2
• Modernized workflow orchestration platform for Apple's Ads Infrastructure group. Built multi-env Airflow infrastructure with automated deployment pipelines and established comprehensive Data Engineering standards. Implemented CI/CD with SonarQube integration, GitHub protection rules, and automated testing gates. Containerized services, onboarded consumer teams, and created knowledge transfer documentation for international team handoff. Reduced manual operational toil by 60%+.

Sep 2018 – Mar 2023 · 4 yrs 6 mos · San Francisco Bay Area

• 🔹 Core Leadership:
• Served as a foundational engineer for the Workflow Platform, focusing on Big Data & Infrastructure.
• Steered as one of the premier engineers, undertaking end-to-end responsibilities from introduction to delegation of multifaceted projects. This involved alignment with cross-functional teams, including security, privacy, and major stakeholders, to achieve holistic company objectives.
• 🔹 Significant Accomplishments:
• Architected the 'Spinner' platform, pivoting from the previous 'Pinball' system, creating a centralized hub for data activities. This encompassed a plethora of features such as lineage tracking, data governance, data orchestration, intuitive visualization, RCA, and more.
• Delivered an enhanced Platform using Airflow, which currently facilitates over 4,500 workflows and manages 60,000+ task instances daily for our expansive user base of 1,500+ internal members.
• Instrumental in achieving a staggering 300% boost in NPS score over a 4-year duration.
• Played a pivotal role in decreasing incidents by 90%, ensuring seamless operations and enhanced user experience.
• Our platform's advancements led to a direct saving of over $5M, due to efficient auto-retirement tooling and streamlined developer operations.
• 🔹 Influencing the Next-Gen of Engineers:
• Spearheaded major design evaluations, cultivating an atmosphere of knowledge transfer, and mentored burgeoning engineers, imparting both technical and soft skills to ensure project success.
• 🔹 Tool & Tech Stack Mastery:
• Proficient in leveraging a wide array of modern big data toolings like Airflow, Kubernetes, Docker, Jenkins, Teletraan, Tereform, Kafka, Spark, and more.
• 🔹 Thought Leadership:
• Shared experiences via blogs and presentations building company brand with external organizations. Have 20+ open source commits to Apache Airflow.

Jul 2018 – Jan 2023 · 4 yrs 6 mos · Lake Forest, California, United States · Hybrid

• I've had the privilege of establishing and growing two pizza parlors alongside my family. As a small business owner, I've gained a comprehensive understanding of the entire operational spectrum
• 🔹 Comprehensive Management: Entrusted with the responsibility of managing a team of over 25 employees, I've overseen every facet of the business – from the careful selection of inventory to the meticulous crafting of our signature recipes and ingredients. This also involves coordinating with third parties, ensuring that the highest quality of products and services are maintained.
• 🔹 Customer-Centric Approach: Our family-centric approach extends to our customer relations. Whether it's addressing inquiries, handling feedback, or ensuring each diner leaves satisfied, the customer experience remains at the heart of our operations.
• 🔹 Financial Acumen: I've delved into the intricate world of accounting, ensuring a seamless financial workflow for the business. This ranges from efficient payroll management to meticulous bookkeeping, ensuring our bottom line is robust and our financial health is in prime condition.
• 🔹 Operational Excellence: As an integral part of business growth, I've pioneered the development of Standard Operating Procedures (SOPs), set up robust HR processes, and facilitated comprehensive employee training. This holistic approach ensures the smooth running of our establishments and the consistent delivery of our renowned pizzas.
• 🔹 Strategic Growth: Being in the realm of a highly competitive market, I've implemented strategies not just to sustain, but to expand. This involves diligent client management and continuously identifying avenues to grow our customer base.
• Owning and operating these pizza locations has been an enlightening journey, teaching me the nuances of every facet of business management. It's a testament to the fact that in business, wearing multiple hats not only broadens one's skill set but also deepens the connection to one's venture.

Jan 2017 – Sep 2018 · 1 yr 8 mos · Oakland, CA

• Proudly wore multiple hats in the realm of data and infrastructure, serving across a myriad of products, ensuring that Pandora's data-driven mission continued with efficiency and innovation.
• 🔹 Apache Airflow Implementation:
• Pioneered the integration of Airflow workflow system, augmenting throughput for teams spanning analysts, engineers, scientists, and products.
• Successfully orchestrated critical workflows, encapsulating listener, catalog, recommendation, and financial data, adhering to SOX/PCI compliances.
• A monumental success was achieved when over 35 production teams adopted the Airflow instance, thus standardizing our operations. We instilled best practices, bolstered security, and enhanced overall cluster submissions.
• 🔹 Presto Integration:
• Instrumental in assimilating Presto into our technical ecosystem, mitigating the overhead of the Hadoop cluster.
• Upon my departure, our Presto cluster thrived with 120+ nodes, resulting in a significant dip in query time
• 🔹 Preferred Version Pipeline:
• Conceived and executed a pivotal pipeline, ensuring listeners received the optimal version of tracks and efficiently streamed rights adjustments, ultimately translating to significant quarterly savings.
• 🔹 SOX/PCI Compliance:
• Aided the Payments teams in transitioning their sensitive workflows to a SOX/PCI compliant Airflow model. Collaborated seamlessly with both internal squads and external audit entities (like EY) to get my design procedures ratified.
• 🔹 Other Key Contributions:
• Engineered the import of third-party advertiser (Adswizz) bid and impression logs via AWS tools, integrating them into our internal clusters.
• Spearheaded on-call responsibilities, decentralizing the analytics organization's on-call burdens.
• Revamped listener data sets, aligning them with the latest product information.
• Engaged deeply with science teams, pioneering the shift from batch jobs to streamlined data via Kafka.
• Played a pivotal role in refining event data, among several other projects.

Jun 2016 – Aug 2016 · 2 mos · Oakland, CA

• Ad Analytics team.
• During my time as an intern on Pandora's Ad Analytics team, I had the invaluable opportunity to contribute to pivotal projects that significantly optimized our big data operations:
• 🔹 Hive Execution Engine Selection: Pioneered a method allowing users to choose their desired Hive execution engine, either via the job URL or directly through the VM config. This advancement ensured a seamless transition between engines without the necessity to alter existing job scripts or files.
• 🔹 Performance Analysis of Tez vs. MapReduce: Delved deep into analyzing the performance metrics of jobs executed using Tez as compared to MapReduce. The results were striking; we observed an average speed-up between 1.5 to 4 times (when comparing CPU time). This not only optimized our operations but paved the way for more efficient data processing.
• 🔹 Production Job Dashboard: Conceptualized and developed an insightful dashboard aimed at providing a comprehensive visual representation of statistics related to production jobs. This tool offers functionalities to group data by various parameters such as VM hostname, job name, task name, etc., allowing users to obtain a granular view of performances across the cluster and identify areas of improvement.
• This internship experience provided a holistic exposure to advanced analytics tools and techniques, reinforcing my passion for data engineering and optimization.

Aug 2016 – Jan 2017 · 5 mos · Los Angeles Metropolitan Area

• Business Intelligence - Data team.
• My contributions spanned three transformative projects that significantly bolstered operational efficiency:
• 🔹 Self-Service Tool for Efficient SQL Script Execution:
• Conceptualized and executed a tool facilitating Product Managers, Business Analysts, and Developers to independently run SQL scripts across their environments. This self-service mechanism enabled seamless alteration of production data without the customary DBA approvals.
• The ripple effect was dramatic: We witnessed a 94% drop in DBA intervention requests, enabling the team to pivot their focus to higher-priority initiatives.
• For the creation of this robust utility, I employed Rundeck, an open-source job scheduler, coupled with intricate Python modules.
• 🔹 Graphical Work Issue Locator for Rocket Fault Detection:
• Designed primarily for the Texas F9 Stage Test unit, this tool found adoption at the Hawthorne HQ and Cape Canaveral site in Florida. Its primary function was to pinpoint rocket anomalies based on physical location, thereby expediting issue resolution prior to Cape-bound shipments.
• Replacing an outdated, sluggish tool, my version, crafted with tSQL, C#, SQLServer, and SSRS, boasted rapid query execution and broader functionality – transforming minutes-long waits into mere seconds.
• 🔹 Machine Learning-Driven Operations Analysis for Industrial Engineering:
• As part of an ongoing initiative, I laid the groundwork for a machine learning project targeting the Production domain. The overarching aim was to unearth similarities across work orders – a historically convoluted challenge for SpaceX.
• My strategy revolved around optimizing clustering and classification rates, and I leveraged Numpy, Scipy in Python, and SQL to realize this goal.

Jan 2016 – Jun 2016 · 5 mos · Oakland, CA, Irvine, CA

• Use various technologies to create web apps for clients. Use a MERN stack (React.js). Integrate client-side and server-side unit testing using the Mocha.js framework on internal tools.

Sep 2015 – Dec 2015 · 3 mos · Berkeley

• Taught the introductory Java and Data Structures course. I held 2 hours of section a week to go over class discussion sheets, than an addition 4 hours a week I held tutoring sessions with small groups to reinforce concepts and answer student questions. I also held office hours and was very active in piazza to reinforce concepts and help debug students' programs.

Sep 2012 – Aug 2014 · 1 yr 11 mos · Berkeley, CA

• Came up with energy saving methods to reduce energy waste throughout campus. Resulted in saving over $100,000 annually on HVAC costs.

May 2014 – Aug 2014 · 3 mos · Singapore, Singapore

• Part of an Alternative Energy Lab focused on Sodium-Ion Battery Research.