May 2025 – Present · 1 yr

• Architected an enterprise-grade AI chatbot ecosystem: 6+ specialized agents built using LangGraph state-machine orchestration, policy-based guardrails, and rate limiting, establishing the core technical foundation for the company’s AI product line.
• Slashed LLM hallucination by 80% via an automated evaluation and regression pipeline in LangSmith with prompt versioning and continuous refinement loops, directly improving production reliability.
• Defined and owned an LLM-powered hyper-personalization engine, persisting long-term user memory in Mem0 backed by Qdrant vector search, extracting structured profiles to serve personalised content across 4 surfaces — articles, videos, polls, and trivia — via a precomputed feed with Bloom filter deduplication, delivering a cohesive per-user experience at scale.
• Spearheaded all cross-cutting technical decisions: model selection, cost-performance tradeoffs, vector search tuning, and conditional agent routing. Functioned as the de facto technical lead across AI and backend teams.
• Launched a live-stream ingestion pipeline for real-time sports broadcast processing. Incoming SRT feeds converted to HLS using AWS MediaLive and Lambda, with segment storage backed by S3, enabling the extraction of downstream highlights at near-real-time latency for live match coverage.

Mar 2022 – Apr 2025 · 3 yrs 1 mo

• Engineered a change data capture pipeline using Debezium to stream database mutations into the caching layer in near-real-
• time, eliminating manual cache invalidation and ensuring read-path consistency across 3+ services.
• Constructed a resumable multi-part video upload system for the highlights product, tracking per-chunk integrity via fingerprint
• hashing in the database ensuring fault-tolerant, idempotent uploads resilient to network interruptions and partial failures.
• Pioneered a social graph service via dual-write denormalization and Redis sorted sets for sub-ms retrieval. Streamlined
• atomic, single-shard query resolution, catalyzing a 25% surge in social discovery and a 12% lift in Day-30 retention.
• Consolidated payment orchestration for 3+ product lines (B2B, D2C, SaaS), integrating Stripe, GCash, and Razorpay.
• Engineered automated post-match ticket reconciliation and SNS/SQS-driven settlement with zero-downtime failover, ensuring
• real-time transaction visibility for global and domestic users.
• Forged a behavioral segmentation engine over user engagement signals, and a gamified in-app currency system
• powering targeted retention campaigns, yielding a 30% improvement in user retention rates and engagement frequency.