Systems Portfolio

A chronological journey through 22+ years of building scalable systems

3

Patents

3

Publications

1

Open Source

99.99%

Availability

100%

AI Adoption

10+

Engineers Mentored

1. LLM-Based Fraud Detection POC

2025/2026 | Expedia Group

Designed and executed proof-of-concept using LLMs to identify fraud in booking transactions by learning from historical analyst decisions. Evaluated ML (XGBoost) vs LLM performance on structured data and created automated decisioning system for analyst-queued transactions.

AI/ML Scalability
ScalabilityProduction-ready fraud detection pipeline on Databricks; targeting 50% analyst queue reduction
AI/MLLocal LLMs (Ollama); UMAP dimensional reduction; vector similarity; prompt engineering; XGBoost benchmarking; Python/Spark
ReliabilityShadow-mode validation against analyst decisions; A/B tested detection accuracy
Cost EffectivenessAutomated fraud detection reducing analyst workload while maintaining high detection rate

2. Multi-Agent LLM System & Agentic AI Innovation

2025 | Expedia Group

Built multi-agent LLM system demonstrating cost-effective reasoning-based problem decomposition using OpenAI GPT (reasoning) and Ollama local models (task execution) with tool/MCP support. Evaluated cutting-edge agentic frameworks including Google A2A and AWS Bedrock AgentCore/Strands SDK.

AI/ML 70%+ Cost Reduction
ScalabilityModular agent architecture; evaluated Google A2A, AWS Bedrock AgentCore, Strands SDK
AI/MLMulti-agent orchestration; reasoning + execution separation; MCP tools; Claude Code; Python
Reliability3 production demos in 3 months; 3 org-wide technical blog posts
Cost EffectivenessHybrid cloud/local model approach reducing API costs 70%+; presented to org-wide Principal Engineer community

3. AI-Driven Developer Productivity Transformation

2025/2026 | Expedia Group

Led organization-wide developer productivity transformation across 8 teams (115 engineers), achieving 100% GitHub Copilot adoption. Created and delivered hands-on training (Python, Java, Kotlin) and established AI Champions program for sustained adoption.

AI/ML Leadership
Scalability115 engineers, 8 teams across multiple locations
AI/ML100% Copilot adoption; AI Champions program; MCP server development; Copilot instruction files
ReliabilityIntroduced ORR template; automated vulnerability remediation with Renovate
Cost Effectiveness20-30% developer productivity gain; 40% faster API/attribute onboarding; 50% reduction in production config changes

4. EG-TNL MCP Server

2025 | Expedia Group

Co-developed MCP (Model Context Protocol) server enabling AI assistants to query live A/B experiment data directly in VS Code through natural language queries—an early adoption of agentic AI patterns.

AI/ML 80% Faster
ScalabilityDockerized for distribution across engineering teams
AI/MLMCP integration with GitHub Copilot; LLM tool-calls; agentic AI architecture
ReliabilityReal-time experiment state queries during development
Cost Effectiveness80% reduction in context-switching time; 3x faster debugging

5. STAR - GenAI RCA Tool

2025 | Expedia Group

Invented STAR (System Telemetry Analyzer), a GenAI-powered root cause analysis tool that automatically collects Datadog metrics and uses LLM prompt engineering for anomaly detection and intelligent root cause analysis during production incidents.

AI/ML Patent Pending
ScalabilityAdopted across all Expedia Group engineering teams for incident automation
AI/MLGenAI/LLM-powered analysis; automated metric correlation; prompt engineering for anomaly detection
ReliabilityRCA delivered within 2 minutes; 80%+ reduction in Mean Time to Know (MTTK)
Cost EffectivenessReplaced hours of manual investigation; reduced on-call engineer cognitive load

6. AWS Operational Readiness Review Process

2024-2025 | Expedia Group

Created Expedia Group's first AWS Operational Readiness Review (ORR) checklist and template through collaborative workshops with AWS and architecture teams, establishing standardized lifecycle governance.

Reliability Process
ScalabilityAdopted across multiple teams organization-wide
ReliabilityMeasurable reduction in incidents after 6 months of adoption
Cost EffectivenessStandardized process reducing operational overhead

7. L3 Incident Command & RCA Leadership

2023-2025 | Expedia Group

Led L3 technical reviews for high-impact incidents affecting revenue-critical services, conducting blameless post-mortems, architectural remediation, PR reviews, and design guidance for affected applications.

Reliability Tier-1
ScalabilityCross-org incident response for Tier-1 revenue-critical services
ReliabilityReduced incident-related revenue impact through architectural fixes
Cost EffectivenessEstablished repeatable RCA process; prevented recurring incidents

8. Clicker - Real-Time Telemetry Platform & SLO Framework

2021-2025 | Expedia Group

Architected and built real-time traveler clickstream telemetry platform processing hundreds of millions of Kafka events into Datadog. Led enterprise SLO transformation improving reliability from 98-99% to 99.9% (front-end) and 99.99% (back-end) for Tier-1 services through standardized SLI/SLOs, Error Budget Policies, and full-stack observability.

99.99% SRE
Scalability100s of millions of Kafka events/day; Kubernetes autoscaling; 100% SLI coverage
Reliability98-99% to 99.9%/99.99% reliability improvement; 100% GraphQL coverage; Istio service mesh metrics
Cost Effectiveness6-10 SLO workshops delivered; 60% Error Budget adoption; strategic partnerships with Google SRE, AWS, Datadog

Error Budget Policy Adoption at Expedia Group

9. ML Traffic Prediction & Capacity Testing Framework

2021-2024 | Expedia Group

Developed ML-based traffic prediction models using Fourier series for peak traffic forecasting across three major brands (Expedia, Hotels.com, Vrbo) handling billions of requests. Built PCT (Peak Capacity Testing) framework and created serverless Traffic Watch data collection system.

AI/ML 95% Accuracy
ScalabilityPredictions for 3 brands handling billions of daily requests; multi-brand PCT framework
AI/MLFourier series time-series forecasting; 95% accuracy (±5%); weekly models for predictive auto-scaling
ReliabilityGameDay exercises for 100% Tier-1 apps; 80% reduction in change-related incidents
Cost EffectivenessTraffic Watch: AWS Lambda/S3 serverless at <$10/year; 4+ years historical data; GitHub Actions automation

Predict It! An Engineer's Journey into Data Science

10. Jarviz - Application Coupling Analyzer

2019-2020 | Expedia Group

Created open-source tool for analyzing application couplings and dependencies.

Open Source
ScalabilityAnalyzes large codebases with complex dependency graphs
ReliabilityIdentifies coupling issues before they cause production problems
Cost EffectivenessOpen-source; reduces technical debt discovery time

Jarviz, Can You Show My Application Couplings?

11. Booking Value Forecast ML Model

2018 | HomeAway/Vrbo

Won first place in company-wide Kaggle competition for Booking Value Forecast, building a predictive model to estimate booking values using historical transaction data.

AI/ML 1st Place
ScalabilityCompetition-winning model for booking value estimation
AI/MLRandom Forest algorithm; feature engineering on booking data; predictive analytics
Cost EffectivenessImproved booking value estimation for business planning

View Kaggle Leaderboard

12. Vrbo Checkout Microservice

2016-2019 | Vrbo/Expedia Group, Austin, TX

Led architecture and development of cloud-native checkout microservice handling mission-critical traveler transactions for web, mobile, and external (Expedia) point-of-sales during AWS cloud migration.

Scalability AWS
ScalabilityMulti-brand, multi-channel checkout serving Vrbo, Expedia, web, iOS, and Android
ReliabilityCloud-native architecture on AWS; resilient transaction handling; circuit breakers
Cost EffectivenessUnified checkout reducing infrastructure costs; enabled multi-brand consolidation

13. Solr-Based Fast Search API

2014-2016 | HomeAway, Austin, TX

Created Solr-based fast search API for internal applications enabling sub-50ms lookups across millions of listed properties.

<50ms Scalability
ScalabilitySub-50ms latency across millions of property listings
ReliabilityClustered Solr with high availability
Cost EffectivenessFast internal lookups improving developer productivity

14. Hadoop HBase Supplier Analytics API

2013-2016 | HomeAway, Austin, TX

Designed scalable supplier analytics reporting API backed by Hadoop HBase, enabling real-time insights for property managers.

Big Data
ScalabilityBig data architecture handling billions of data points
ReliabilityDistributed HBase cluster with replication
Cost EffectivenessReal-time insights enabling data-driven property management decisions

15. Partner Dashboard Platform

2013-2016 | HomeAway, Austin, TX

Led team building scalable web-app solutions for small-scale property owners and large-scale property managers.

Scalability
ScalabilityServed thousands of suppliers managing millions of property listings
ReliabilityProduction-critical platform for property manager operations
Cost EffectivenessSelf-service platform reducing support overhead

16. JavaScript Ad Delivery Framework

2011-2013 | Bazaarvoice, Austin, TX

Built zero-config JavaScript framework for embedding UGC-powered ads on publisher websites.

ML Integration
ScalabilityDeployed across thousands of publisher sites
AI/MLIntegrated with ML-generated ad snippets
ReliabilityClient-side resilience with fallback mechanisms
Cost EffectivenessZero-config deployment reducing integration costs

17. UGC Ad Snippet REST API

2009-2013 | Bazaarvoice, Austin, TX

Built highly scalable REST API delivering ad snippets auto-generated from user-generated content (reviews, ratings) using NLP and machine learning.

AI/ML Patents
ScalabilityHandled millions of UGC records, real-time ad serving at scale
AI/MLNLP-based snippet extraction from reviews, ML-powered content relevance
ReliabilityHigh-availability ad delivery system with strict latency SLAs
Cost EffectivenessAutomated snippet generation replacing manual curation

18. FastTrack - Order Management System

2006-2009 | Toppan Photomask, Round Rock, TX

Built web-based user management and order management system for one of the largest photolithography service providers in the US.

Enterprise
ScalabilityEnterprise system supporting high-volume semiconductor manufacturing orders
ReliabilityFull-stack enterprise application with production-critical workflows
Cost EffectivenessStreamlined order processing reducing manual overhead

19. EID - Student Authentication System

2006 | University of Texas at Austin

Contributed to Java-based online student authentication and authorization system serving UT Austin's student population.

Security
ScalabilityUniversity-wide system serving 50,000+ students
ReliabilitySecurity-focused identity management with high availability requirements
Cost EffectivenessCentralized authentication reducing duplicate identity systems

20. Flight Reservation Module (TravelBox)

2003-2005 | CodeGen.IT, Sri Lanka

Built flight reservation module for TravelBox tour operating system, an enterprise travel technology platform.

Enterprise
ScalabilityEnterprise-grade booking system designed for multi-tenant tour operators
ReliabilityMission-critical reservation flows handling real-time inventory
Cost EffectivenessProductized module enabling faster client deployments