* Led development of a Retrieval-Augmented Generation (RAG) system that reduced ad sales segment search time by >80%, replacing manual spreadsheet processes and accelerating decision-making for 20+ sales stakeholders.

* Architected a hybrid search pipeline combining SentenceTransformers and BM25 to power semantic + keyword search over CSVs with 50+ columns, improving data retrieval accuracy and usability by ~70%.

* Implemented OpenAI-based tool-call routing, enabling dynamic execution of semantic or structured queries with a ~95% success rate, and improved user experience by reducing back-and-forth clarification prompts by >50%.

* Oversaw full project lifecycle, leveraging Jira and Confluence, including architecture design, sprint planning, mentoring 2 junior developers, and coordinating with 3+ cross-functional teams (data science, product, business ops).

* Built a finance RAG-powered reporting platform with a React frontend interface and backend document generator with Python FastAPI.

* Automated monthly and quarterly financial reporting, reducing manual effort by ~90% and cutting report turnaround time from 1–2 days to under 10 minutes via LLM-based summarization and SQL-driven data retrieval with dynamic output generation.

* Originated report generation engine with zero-to-one development leveraging AWS Lambda, ECS, SQS, and S3, using the orchestrator pattern to manage sequential and parallel operations, optimizing concurrency, CPU and memory usage, and lambda cold start times.

* Acted as team lead during lead absence, managing Jira boards and agile ceremonies, balancing 3 developer’s workloads, achieving sprint goals, maintaining Confluence documentation, providing code review, meeting with senior management and product teams.

* Reduced generation time by 45% on a proprietary PDF report library in Typescript built upon PDFKit for custom creation of CRE data reports relating to properties, areas, comparisons and more, to meet client-driven design, accessibility, and product requirements.

* Created Jest snapshot tests for difficult to test PDF components and optimized coverage of helper functions in the engine to 75%.

* Utilized Datadog for observability tooling on report generation engine to monitor usage and performance metrics.

* Implemented over 40 visualizations using Highcharts and table presentations with AGGrid of proprietary risk exposure data translated into key real estate performance metrics for use in frontend React app and custom generated PDF reports for investor use.

* Utilized DBT (Data Build Tool) to automate and streamline data transformation processes, increasing efficiency and reliability of data.

* Appointed Vite as build tool to achieve over 50% decrease in build time compared to webpack, expediting development time.

* Refactored large scale existing architecture into a monorepo using nx to facilitate code sharing and accelerate development cycles.

* Used test-driven React development to create a new climate risk feature presenting data for thousands of commercial properties and geographies alongside traditional performance metrics for Moody’s Analytics CRE (Commercial Real Estate) REIS investor tool.

* Created Jest snapshot tests for difficult to test PDF components and optimized coverage of helper functions in the engine to 75%.

* Provided mentorship, technical guidance and support to junior engineers and interns, and participated in the Moody’s Foundation volunteer initiative to provide resume feedback and mock interview experience to 2 new internship applicants and college students.

* Harnessed Google Maps API to create map views with up to 50 custom markers and shape overlays based on user selected data.

* Designed with SASS to leverage advanced features such as variables and nesting in the frontend application components.