Experienced Backend Engineer with 8+ years of experience in scalable system design, microservices architecture. Skilled in Java, C#, and C++, with expertise in SQL/NoSQL databases, CI/CD automation, performance optimization, and cross-service integration. Collaborative problem-solver with a strong track record of delivering high-impact engineering outcomes in fast-paced environments.
1. Hybrid OCR for Office and Image Files
Situation: The OCR feature could not handle hybrid Office files (text + image), leading to billing issues and incomplete document processing.
Task: Deliver a production-ready solution that correctly processes hybrid documents and avoids double billing.
Action: Took end-to-end ownership of the Hybrid OCR rollout, collaborated with Content Store, MeTA, and Vroom teams, created new persistent SPO metadata fields for caching, and led a successful security review.
Result: Achieved 100% production rollout; feature enabled 93% cost savings per 1,000 images and expanded OCR coverage for premium offerings.
2. Cosmic Migration of SDE Microservice
Situation: The global initiative to move services from RDFE to Cosmic was high-priority and time-sensitive. Cosmic is an internal wrapper over Kubernetes and Docker.
Task: Migrate SDE service with full feature parity and reliability.
Action: Mapped RDFE components to Cosmic equivalents, created dynamic pod configs, fixed memory leaks, implemented partition-level compute config, and led knowledge transfers.
Result: Achieved 100% production rollout of SDE in Cosmic; recognized for end-to-end ownership and unblocker role across teams.
3. OCR Backfilling Design and Implementation
Situation: Customers needed OCR for historical files, not just new uploads.
Task: Enable accurate and efficient backfilling of OCR without duplicate billing.
Action: Led design discussions, validated billing logic, resolved race conditions, and created new telemetry types to distinguish backfill traffic.
Result: Broadened OCR applicability, unlocked revenue from archival content, and set foundation for reliable backfill architecture.
4. Resource Isolation for MeTA Infrastructure
Situation: Organizational push to isolate Azure resources for enhanced security.
Task: Implement infrastructure-level isolation in MeTA.
Action: Extended EV2 rollout specs, secured ACS/KeyVault/Redis resources, engaged in cross-team security reviews, and documented deployment flows.
Result: Delivered a compliant, hardened infrastructure within tight deadlines, praised for completeness and reuse by partner teams.
5. BlobDecryptionError Reduction in Tanagra
Situation: Persistent BlobDecryptionErrors (400K/month) due to key mismatches disrupted Tanagra reliability.
Task: Identify the root cause and implement a robust fix.
Action: Diagnosed encryption key mismatches between CPS and MeTA, implemented fixes, and validated in production.
Result: Reduced errors by 99% (from 400K to ~3K/month), improving Tanagra reliability from 98.91% to 99.26%.
6. XorHash Force Recompute Optimization
Situation: Redundant force recomputation of XorHash caused latency spikes (7–10s) in the CPS pipeline.
Task: Eliminate unnecessary recomputations to improve performance.
Action: Refactored GetLightInfo logic, added DocumentTag-based BSN-QXH checks to bypass expensive recomputes for unchanged files.
Result: Reduced 689K forceGenerateXorHash calls/month, significantly improving latency and saving compute resources.
7. Tanagra Local Test Environment Enablement
Situation: Developers lacked a way to validate full pipeline behavior locally, delaying feedback loops.
Task: Enable local testing of Tanagra pipeline components.
Action: Connected CPS SPO sandbox with a local MeTA Docker container, bypassed service buses, and validated file fetch and profile module execution.
Result: Empowered developers to test real pipeline logic locally, reducing iteration cycles and increasing pre-dogfood deployment confidence.
8. Leadership and Team Enablement
Situation: Team members lacked deep MeTA/CPS expertise, slowing progress.
Task: Transfer knowledge and build team self-sufficiency.
Action: Mentored multiple teammates, conducted onboarding sessions, clarified Tanagra/CPS/MeTA logs, and provided critical debugging aids.
Result: Improved team velocity, reduced dependency on external SMEs, and increased PR review quality across time zones.
Java, C, C#, PowerShell