Jackie Zhai

1. Bangladesh Card System Integration Successfully integrated the card system in Bangladesh, with a daily transaction volume of approximately USD 10,000. 2. Vietnam Card System Integration Completed the card system integration in Vietnam, supporting a daily transaction volume of approximately USD 60,000. 3. AWS Infrastructure Setup Deployed and managed AWS infrastructure, including MySQL, Redis, Zookeeper, ECR, and implemented containerized deployment using CodePipeline. 4. Bangladesh Project Responsibilities Led the team in the development of the following modules: * Payin and Payout processing * Reconciliation * Manual order confirmation * Fee rate management * Balance management * Card provider management * Card configuration management * Monitoring and alerting * Card transaction history management 5. Vietnam Project Responsibilities Led the team in completing: * Payin and fund consolidation (aggregation) development * Risk control module for collection accounts * Fund consolidation algorithm development * Reconciliation * Manual order confirmation * Monitoring and alerting * Bank transaction history management

1. Designed and implemented 7 SLA agreements to improve supplier integration response time by 30%. 2. Created SOPs that reduced integration errors by 25%. 3. Led a team to complete integration of NI POS and e-commerce payment system, cutting integration time by 40%. 4. Developed automated top-up via Central Bank to PayBy accounts, reducing manual operations by 50%. 5. Implemented automated salary payment system for WPS via FTS, cutting processing time by 70%. 6. Built real-time FX rate alert system, boosting user engagement by 20%. 7. Designed MoneyGram failure email alert & tracking system, improving customer satisfaction by 15%. 8. Developed coupon code feature, increasing transaction volume by 25% during campaigns.

Led end-to-end technical integration and solution design for the Brazil payment ecosystem, including: 1. Integration with Brazilian Credit Vendor QI 2. Integration with Brazilian Merchant VTEX 3. Integration with Brazilian Payment Channel BePay 4. Integration with Merchant Express Software (LightSpeed Ecosystem)

Summary Led backend system stability for 99Pay, DiDi’s financial app in Brazil, handling rapid user growth and high transaction volume. Focused on vendor management, performance tuning, and country-specific integration for Brazil and Mexico. Key Contributions & Achievements 1. Integrated Mexico wallet system, supporting 5,500+ daily mobile recharges, 200+ utility bill payments, and 30+ concierge card transactions. 2. Improved biller 6900 & 37 success rate from 20% to over 99%, resolving 70% of utility bill traffic issues in Mexico. 3. Reduced BePay peak traffic load by 60% using Redis and Apollo-based dynamic throttling. 4. Implemented PIX registration optimizations, lowering Brazil-side throttling risk by 20%. 5. Shortened wallet reward processing duration by 50% during peak load periods. 6. Led BePay AWS migration and domain switchover, unifying under 99Pay branding. 7. Established vendor outage SOP and full incident workflow for Matera platform stability.

1. Rule Engine Data Analytics - Visualized rule users: users with assigned rules and users with executed rules via line charts. - Total number of rules: displayed using both line charts and pie charts. - Active rules: analyzed and visualized through line and pie charts. - Rule executions: statistical line and pie chart views of execution count. - Execution results: success/failure rates represented in line and pie charts. - Execution device analytics: success rates by rule type and product model. 2. Firmware Data Channel - Developed OpenResty module to retrieve firmware data. - Processed and forwarded data to Kafka using Lua scripts. - Built local test environment with Kafka, Elasticsearch, and ES-Head containers. - Independently developed a Spring Boot system integrating Kafka listener to consume data and store it in Elasticsearch. - Configured index mapping and resolved date type compatibility issues in Elasticsearch. 3. Event Tracking Statistics - Used Waterdrop and Spark cluster to consume Kafka data and define input, filter, and output. - Ingested data into TiDB with hierarchical modeling (ODS, DWS, ADS layers). - Generated metrics on device activation, device binding, user retention duration, median and average statistics. 4. Azkaban Scheduling System - Implemented scheduled tasks using Azkaban to trigger Python jobs, performing ODS-level calculations and writing results into ADS. 5. Go-based BI System - Developed permission management module for the BI system using Golang.

Project Overview Led backend development for a Smart Home system with key features including visitor reservation, resident verification, community services, property payments, maintenance requests, and more. The system consisted of a mobile app (Jinshan Cloud Smart Home) and a web portal, utilizing technologies such as Spring Cloud, MyBatis, RabbitMQ, and JPush, with integration of Hikvision devices and community data. Key Responsibilities & Achievements 1. Developed data analytics modules for residents (by gender and age), visitors (by type), door locks (usage stats), and device alarms. 2. Implemented synchronization of community, building, unit, personnel, and device data with Hikvision platform. 3. Subscribed and processed Kafka topics from Hikvision using factory pattern with manual offset control. 4. Built work order module: residents submit via app, property managers handle via web; supported multi-space isolation and producer-consumer pattern. 5. Implemented convenient contact management for adding, deleting, and updating community service numbers. 6. Integrated visitor appointment system: created reservations via Hikvision API and generated QR codes for entry. 7. Displayed visitor logs on web and mobile, with support for record deletion. 8. Developed async callback service for video intercom: server triggers JPush notifications to mobile clients via thread pool. 9. Handled Hikvision door unlock callbacks asynchronously to ensure real-time responses. 10. Led the successful integration of 'Wuhan Youth City' project into the existing Aihouse system, ensuring seamless data and feature alignment

Project Overview Aihouse is a smart living solution platform tailored for residential communities and commercial properties. It offers a full-stack ecosystem that includes resident service systems, device control, scene-based automation, voice interaction (via Xiaomi XiaoAI), and WeChat mini programs for both B2B and B2C scenarios. During my time at Aihouse, I played a key role in evolving the platform from a traditional monolithic system into a scalable microservices architecture. My work focused on enhancing system reliability, performance, and real-time interaction, while supporting both smart home users (residents, hotel guests) and operational stakeholders (property managers, merchants). The platform supported: 1. Voice-activated services (e.g., controlling non-Xiaomi devices via NLP) 2. Smart IoT device orchestration (e.g., locks, meters, cameras) 3. Data analytics for operations and user engagement 4. Edge computing scenarios for real-time lighting control based on camera ROI detection 5. ToC & ToB services via WeChat mini programs with secure authentication and order fulfillment This comprehensive smart living platform was deployed in residential communities, smart hotels, and apartment complexes, providing a seamless digital experience for thousands of end users and property stakeholders.

Business Context The project focused on building a productized big data reporting system for JD.com’s advertising ecosystem. It processed massive-scale user behavior data (tens of billions of records) generated through third-party DSP platforms such as Pinyou, Zhimai, and Ruixuan. The entire data pipeline integrated Kafka, Hive, Spark, MapReduce, and Oozie for real-time and batch processing, with results exposed through dashboards and APIs to business users and marketing systems. Key Responsibilities & Achievements 1. Wrote Hive SQL scripts to compute metrics like mid-touch, final-touch, views, cart additions, product follows, and orders using functions like mapjoin, row_number() over, base64/unbase64, locate, date_sub, from_unixtime. 2. Replaced static Oozie Hive job triggers with dynamic Spring Boot scheduling system, supporting parallel multi-source (Pinyou, Zhizhi, Ruixuan) execution. Used fork nodes and Hadoop distcp to sync results to cloud/MySQL. Reduced scheduling latency by 30%. 3. Applied ALS and FPGrowth algorithms via Spark MLlib to recommend personalized products, boosting user engagement by 25%. 4. Built a web-based UI to manually manage Oozie jobs (kill, restart, trigger), extract report data, and generate audience segments. Used Spark SQL for data aggregation. 5. Developed components for report data retrieval and result callback integration. 6. Used Spring Boot + reflection to read Hive tables in parallel, and created product/category tables. Exported results via Scala to OSS after storing in HDFS via Oozie. 7. Implemented lightweight validation workflows using embedded H2 database. 8. Designed Oozie DAG scheduler to monitor and alert on critical computation node status. 9. Prepared compute resources and extracted precise audience packages with auto-cleanup. 10. Consumed Kafka ad event streams using Flink and persisted to HDFS. 11. Designed and delivered interactive BI dashboards using ECharts for marketing and analytics teams.

Project Overview StageDB is a data processing and integration platform designed as an ETL system. It handled files received from Ariba, including contracts, POs, GRs, users, and vendor data. The system parsed, transformed, and persisted the data into a PostgreSQL database. Further enrichment was performed via stored procedures, and the final output was distributed to downstream systems such as ECC and ITPR. Quartz was used extensively to schedule and orchestrate all ETL tasks. Key Responsibilities • Retrieved and parsed ZIP attachments from emails using JavaMail based on subject filtering. • Processed CSV and XLS files from specified Linux directories. • Pulled and parsed XML files from FTP servers. • Delivered processed data to downstream systems via email or FTP. • Monitored critical data (contracts, PO GR, invoices) and triggered alerts on anomalies. • Handled full and incremental updates for Ariba user and vendor data. • Developed authenticated web service interfaces to expose contract data to external systems. • Invoked multi-stage stored procedures (ST, IDL, PCL, RDL) via Java for data processing. • Designed multithreaded components to process incremental vendor data and supplement contracts. • Managed incremental updates for various data types using stored procedures. • Implemented robust exception handling around SQL procedures to support historical data processing.