Experience

Promoted from intern to Associate Backend Developer, now a backend owner across major slices of the platform.

Backend engineer on Vonage “Journey Architect” — the execution engine of a workflow-orchestration platform that runs complex multi-channel user journeys (WhatsApp, SMS, RCS, webhooks), built on Temporal + GCP before a strategic migration to an event-driven AWS + RabbitMQ model.

Orchestration engine (Temporal)

• Engineered core orchestration nodes — the asynchronous “Wait for Action / Event” steps — mapping complex DSL structures into executable Temporal worker activities.
• Architected the Step Registry for delay and wait-based events, enabling dynamic workflow execution.
• Built worker schemas and step definitions so new node types onboard cleanly.

GCP → AWS, event-driven re-architecture

• Key backend driver of the migration from GCP to an AWS + RabbitMQ event-driven model.
• Rebuilt the “Delay”, “Wait for Engagement”, and “Wait for Event” workers from synchronous processing to message-queue execution.
• Migrated and optimized the analytics processing with zero data loss through the cutover.

DSL & execution logic

• Extended the engine to evaluate multiple conditions per branch, and to process OR / boolean logic in Conditional and Wait-for-Event steps.
• Added, defined, and processed new event triggers in the orchestration engine.

Analytics, observability & security

• Co-built journey-execution analytics, then independently built a step-wise, node-by-node metrics system.
• Shipped Debug Logs for real-time workflow observability in the UI.
• Built centralized Secrets Configuration APIs to securely store and pass third-party auth parameters.

Rotated across two backend teams during the internship — first hardening the Requirement AI platform, then modernizing data architecture on the EIT team.

Requirement AI — backend & cloud

• Built and maintained core endpoints, including the convert_to_pdf document-conversion pipeline and its utilities.
• Integrated Google Drive and Google Cloud Storage for artifact retrieval (get-artifact, find_shared_artifact), and surfaced artifact_title in check_request for better process tracking.
• Added fuzzy search for shared artifacts (by title and ID), making them far easier to find.
• Introduced prompt versioning to track changes to AI system prompts over time.
• Removed hardcoded secrets, closed security holes, and refactored across branches for maintainability.
• Built the Prompt Management section of the admin dashboard.

EIT modernization — data & scale

• Built a scheduler that auto-triggers and processes documents, with overwrite support and manual reprocessing.
• Migrated MongoDB collections to a versioned schema that preserved history while cutting storage — and DB cost.
• Rewrote hot API paths as MongoDB aggregation pipelines, cutting response times sharply.
• Owned the backend for the new User Dashboard, with fast paginated loading.
• Tuned LLM prompts for better output, and reorganized Azure storage for a cleaner client experience.

• Contributed to the company’s core product framework and led the build of its flagship demo product, Climb, on a custom in-house stack.
• Co-designed and engineered an AI product on Apple ML-4M, building a scalable ML pipeline capable of real-time data processing.
• Improved security and performance by integrating OIDC-based authentication and optimizing feature classes — reducing login latency.