Zoraiz

Evaluating search relevance and AI-generated responses against project guidelines for quality, accuracy, and usefulness.

• Evaluated search relevance across assigned TELUS rating tasks by analyzing user intent, result accuracy, page quality, and usefulness against project guidelines.
• Reviewed AI-generated text responses across daily evaluation tasks by rating factual accuracy, instruction following, completeness, clarity, and natural language quality.
• Classified search and content quality issues such as irrelevant results, misleading information, unsafe content, and poor query matching.

Building a Microsoft Fabric medallion lakehouse for IBM Db2 data — from ADF ingestion through Bronze, Silver, and Gold analytical layers.

• Engineered a Microsoft Fabric medallion architecture for IBM Db2 data by using Azure Data Factory to ingest source tables into OneLake, enabling structured Bronze, Silver, and Gold lakehouse processing.
• Built metadata-driven ADF pipelines using parameterized source queries, dynamic sink paths, table configuration metadata, and watermark-based logic, reducing manual pipeline duplication across source tables.
• Landed raw Db2 extracts into the Bronze layer in Parquet format, preserving source-aligned data structures while enabling scalable downstream transformation in Microsoft Fabric.
• Transformed Bronze data into Silver Delta tables by applying schema standardization, column normalization, data type enforcement, deduplication, null handling, and incremental merge logic to improve data quality and consistency.
• Developed Gold analytical tables from curated Silver datasets using joins, aggregations, business rules, and dimensional modeling patterns, enabling Power BI-ready reporting outputs.
• Implemented audit logging and table-level error handling across ingestion and transformation workflows, improving observability into load status, watermark values, execution timestamps, and failure details.

Architected and scaled a real-time multiplayer game server and production infrastructure for a live player base of 17,000+ unique users.

• Architected and maintained a real-time multiplayer game server in Lua, implementing core gameplay systems (clans, duels, combat) using event-driven client–server architecture.
• Designed a map streaming system using spatial partitioning and interest management, reducing bandwidth usage and improving client performance through batched network events.
• Integrated Discord API via a Node.js service to enable real-time communication and event synchronization between in-game players and Discord channels.
• Developed custom real-time rendering and UI systems (radar, HUD, dynamic overlays), optimized for low-latency updates and smooth client performance.
• Deployed and managed production infrastructure on a Linux VPS (OVH), configuring Nginx for asset delivery and supporting monetization through in-game purchases.
• Scaled and supported a live player base of 17,000+ unique users, ensuring stability, performance, and continuous feature delivery.

Engineered and optimized a volumetric live-streaming system in Node.js, documenting findings in a research paper.

• Engineered a volumetric live-streaming system in Node.js, increasing frame rate from ~30 FPS to ~38 FPS (~25% improvement) through performance benchmarking and targeted optimizations.
• Identified and resolved performance bottlenecks, improving overall system efficiency.
• Leveraged Docker containers for platform independence, ensuring seamless deployment across environments.
• Designed a modular architecture, enabling scalability and future enhancements.
• Authored a research paper documenting findings and improvements.

Led a machine learning project analyzing student feedback to generate data-driven course recommendations.

• Developed a machine learning model in Python using decision trees and information gain to analyze student feedback.
• Generated data-driven course recommendations, improving decision-making for students.
• Enhanced student satisfaction by providing personalized insights based on feedback analysis.