Seattle, Washington, United States
Experienced Software Developer with a demonstrated history of delivering multiple projects and features in the software industry. Skilled in core java technologies, back-end development, object oriented design, automation scripting and Oracle Database and Clusterware operations. Strong engineering professional with a Master of Science (MS) in Computer Science from Columbia University in the City of New York.
Technical lead of the AI engine powering Indeed's Sourcing Assistant, which automates recruiter candidate search using LLMs and multi-strategy retrieval. Launched publicly in 2026; generally available across all employers. (Jan 2026 – Present) Architected a multi-agent LLM orchestration system (LangGraph) with iterative retrieval, multi-pass scoring, and reflection loops — 300% improvement in optimal candidate matches over the prior single-model approach Led GPT-5 model migration: systematic offline evaluation across 5 model/reasoning configurations, delivering a data-driven recommendation to use a smaller model for specific prompts at 1/3 latency with acceptable quality trade-offs Designed a retrieval strategy using LLM-refined Lucene queries that contributed 31% of top candidates; removing it degrades employer approval rate by 2.3 percentage points Drove location-aware LLM evaluation as a new architectural capability — 50% fewer empty results, ~39% fewer low-result searches Designed Quick Sourcing, a low-latency pipeline variant completing in 30s vs. 3.8min (87% reduction) with quality parity — live in production Implemented Responsible AI guardrails across all LLM prompts — anti-discrimination instructions, prompt injection mitigation, PII redaction — prerequisites for scaling to 25,000+ users Core engineer on Indeed's first AI recruiting assistant. Built the candidate sourcing pipeline from prototype through beta across 32 employers. (Dec 2024 – Dec 2025) Built the execution layer orchestrating candidate retrieval via fan-out parallelism and LLM-based evaluation with structured outputs (Pydantic) Designed weighted match criteria to prioritize explicit recruiter requirements over job-derived attributes in LLM prompts; added hard filtering for must-have criteria Delivered multi-location support with per-location batching and merge logic that maintained pagination integrity despite single-location upstream API constraints
Event-driven services for candidate lifecycle data. Led the Candidate Timeline Service end-to-end: requirements, design, GraphQL API (Apollo Federation), DynamoDB storage, and cross-team publisher onboarding Built the Budget Recommendation Service from scratch, integrating ML models for job performance predictions Designed Kafka Streams pipelines to process employer/jobseeker email events and derive employer interest signals Reached Trainer status for technical interviews — the highest interviewer level — conducting 75–90+ sessions across SWE, SRE, Data Science, and Manager roles
Worked on new feature in Oracle Grid Infrastructure called Rapid Home Provisioning (RHP) which enables rapid installation and patching of Oracle Database and Clusterware from a centralized RHP server: - Expanded RHP to automate patching of Oracle Database Appliance (ODA) and Exadata machines using core Java technologies, Expect scripts and Java SSH (JSCh) libraries -Developed script that automates installation of Oracle Grid Infrastructure and starts RHP server to be used for cloud deployments - Implemented feature to configure posix groups and assign Oracle Database roles to them during installation - Implemented functionality to provision software on remote machines with different operating systems than RHP server
Developed an application to display a 3D model on Google glass whose orientation is controlled using touch gestures and voice commands Development work in Glass Development Kit - an add-on to Android SDK development tool
Developed a Java Application to produce a report for audit purposes from requirement analysis to production deployment Performed PL/SQL research and testing for Automation of Database Deployment, with emphasis on running SQL dynamically from a PL/SQL anonymous block Practiced Agile/Scrum Methodologies - attended Sprint planning, Sprint review, Daily stand-up meetings
Project: Development of Generic tool for creating software driver for Power Management Software Programming in C#