Tech Support / DevRel Engineer at Orderly Networks

Architected and built a Python-based MySQL ORM automation tool that reduced manual operations by 80% and saved more than 20 hours weekly across over 50 marketing and business team members. Provided technical support for over 500 enterprise clients on high-frequency trading infrastructure, achieving a 99.5% issue resolution rate and optimizing $10M+ daily trading volume. Designed a real-time multi-dimensional monitoring system using Prometheus and Grafana, reducing system downtime by 60% and enabling proactive issue detection for trading infrastructure handling over 100K transactions per second. Led developer relations initiatives that tripled developer adoption and increased API integration from 15% to 45%. Built an automated testing framework covering REST APIs, WebSocket connections, and SDK interfaces with 95% test coverage, reducing production bugs by 40%.

Founder & ML Engineer at 02AI

Founded and developed an LLM-powered research assistant serving more than 200 bioinformatics researchers, improving academic paper analysis accuracy by 85% through advanced Retrieval-Augmented Generation (RAG) implementation. Architected a scalable RAG system with Model Context Protocol server in Python and FastAPI, processing over 10,000 research papers with sub-2 second query response times. Designed a hybrid memory architecture with Redis and PostgreSQL for long-term LLM conversation context retention across more than 1000 sessions. Built a multimodal scientific data interpreter handling images, tables, and charts with 90% accuracy integrating multiple LLM providers for enhanced reliability. Implemented an enterprise-grade orchestration system managing concurrent LLM requests and tool calling, reducing data analysis workflow times from hours to minutes.

Software Development Engineer at Microsoft

Optimized cloud service infrastructure memory usage by 25% through efficient data structure implementation and caching strategies, saving over $50K annually for services handling more than 1 million daily requests. Developed and maintained 15+ automated monitoring tests using C# and Azure DevOps, achieving 99.9% service uptime and reducing mean time to detection from 30 minutes to 5 minutes. Implemented responsive frontend features using TypeScript, React, and debugging tools, improving user experience metrics by 30% and reducing customer-reported UI issues by 45%. Resolved more than 200 customer support tickets through detailed log analysis and root cause investigations, maintaining a 4.8/5 customer satisfaction score and reducing average resolution time from 48 to 12 hours.

Research Assistant & Full Stack Developer at Doxey Lab at University of Waterloo

Built and deployed a gene visualization web application serving over 500 global researchers with 99.5% uptime using Flask, MongoDB, MySQL, and Docker on Google Cloud. Optimized billion-row biological database (GTDB) queries by 70% through indexing and query restructuring, reducing analysis times from hours to minutes. Designed a novel genomic clustering algorithm combining hierarchical clustering with Smith-Waterman distance metrics via dynamic programming, improving accuracy by 25% compared to existing methods. Collaborated with an 8-person interdisciplinary research team using Agile, contributing to 3 peer-reviewed publications and 2 international conference presentations.

Machine Learning Developer Intern at ByteDance AI Lab, TikTok Security Team

Developed a large-scale fraud detection system using Hadoop and Spark processing over 100 million daily user interactions, achieving 92% accuracy and reducing false positives by 35%. Implemented LSTM-based sequence analysis and FP-Growth algorithms for behavioral pattern recognition, detecting more than 15,000 fraudulent accounts and protecting platform value exceeding $2 million. Built an automated web crawler with Selenium and Python to monitor over 50 underground market websites, providing intelligence that led to over 200 account takedowns and enhanced platform security.

Computer Vision Intern at RetailEye Intelligence Ltd

Developed a CNN-based Optical Character Recognition system achieving 94% accuracy on retail product labels, processing over 50,000 images daily and reducing manual data entry costs by 60%. Built a scalable image preprocessing pipeline using OpenCV and Python to optimize data quality for machine learning models and reduced model training time by 40% with automated augmentation techniques.

Tech Support / DevRel Engineer at Orderly Networks

Architected and built Python-based MySQL ORM automation tool used by 50+ marketing and business team members, cutting manual effort by 80% and saving over 20 hours weekly. Supported 500+ enterprise clients on high-frequency trading infrastructure with a 99.5% resolution rate, contributing to daily trading volume optimization over $10M. Designed real-time multidimensional monitoring system with Prometheus and Grafana, reducing downtime by 60%. Led developer relations initiatives increasing developer adoption by 300% and API integration rate from 15% to 45%. Developed automated testing framework covering REST API, WebSocket, and SDKs with 95% coverage, reducing production bugs by 40%.

Founder & ML Engineer at 02AI

Founded and developed an LLM-powered research assistant serving over 200 bioinformatics researchers, improving academic paper analysis accuracy by 85% through advanced Retrieval-Augmented Generation (RAG) implementations. Architected scalable RAG system processing 10K+ research papers with sub-2 second query response times using Python, FastAPI, and vector databases. Designed hybrid memory architecture enabling context retention across 1000+ sessions using Redis and PostgreSQL. Built a multimodal scientific data interpreter with 90% accuracy integrating multiple LLM providers (OpenAI, Anthropic, Google). Implemented enterprise-grade orchestration system to manage concurrent LLM requests and tool calling, reducing data analysis workflows from hours to minutes.

Software Development Engineer at Microsoft

Optimized memory usage across cloud service infrastructure by 25%, saving over $50K annually for services handling more than 1 million daily requests. Developed and maintained 15+ automated monitoring tests with C# and Azure DevOps, achieving 99.9% service uptime and lowering mean time to detection from 30 minutes to 5 minutes. Implemented responsive frontend features using TypeScript and React, improving user experience metrics by 30%, and reducing customer-reported UI issues by 45%. Resolved more than 200 customer support tickets via log analysis and root cause investigation, maintaining a 4.8/5 customer satisfaction score and cutting average resolution time from 48 to 12 hours.

Research Assistant & Full Stack Developer at Doxey Lab at University of Waterloo

Built and deployed a gene visualization web application serving over 500 researchers globally with 99.5% uptime using Flask, MongoDB, MySQL, Docker, and Google Cloud. Optimized GTDB biological database queries that handle billion-row datasets with a 70% performance improvement through indexing and query restructuring, reducing analysis time from hours to minutes. Designed a novel genomic clustering algorithm combining hierarchical clustering and Smith-Waterman distance metrics, implemented through dynamic programming—resulting in a 25% accuracy improvement. Collaborated within an 8-person interdisciplinary team using Agile and contributed to 3 peer-reviewed publications and 2 international conference presentations.

Machine Learning Developer Intern at ByteDance AI Lab, TikTok Security Team

Developed a large-scale fraud detection system processing over 100 million daily user interactions with Hadoop and Spark, achieving 92% accuracy and reducing false positives by 35%. Implemented LSTM sequence analysis and FP-Growth algorithms to recognize behavioral patterns, detecting over 15,000 fraudulent accounts and protecting platform value exceeding $2M. Built an automated web crawler using Selenium and Python monitoring 50+ underground market websites, providing intelligence leading to more than 200 account takedowns, boosting platform security.

Computer Vision Intern at RetailEye Intelligence Ltd

Developed a CNN-based Optical Character Recognition system achieving 94% accuracy on retail product labels, processing over 50,000 images daily and reducing manual data entry costs by 60%. Built a scalable image preprocessing pipeline using OpenCV and Python, optimizing data quality and reducing training time by 40% through automated augmentation techniques.

Tech Support / DevRel Engineer at Orderly Networks

Led development of a Python-based MySQL ORM automation tool serving 50+ teams; reduced manual operations by 80% and saved 20+ hours weekly. Provided technical support for 500+ enterprise clients on high-frequency trading infrastructure, achieving 99.5% issue resolution and contributing to $10M+ daily trading volume optimization. Designed and deployed a real-time, multi-dimensional monitoring system with Prometheus/Grafana, reducing downtime by 60% for 100K+ TPS. Led developer relations including API docs, GitHub SDKs, and community engagement, boosting developer adoption by 300% and API integration from 15% to 45%. Built an automated testing framework covering REST APIs, WebSockets, and SDK interfaces, achieving 95% test coverage and reducing production bugs by 40%.

Founder & ML Engineer at 02AI

Founded and developed an LLM-powered research assistant serving 200+ bioinformatics researchers; achieved 85% accuracy improvement in academic paper analysis via RAG. Architected scalable RAG system with Model Context Protocol (MCP) server using Python, FastAPI, and vector databases; processed 10K+ papers with sub-2s query responses. Developed frontend components using Vue.js and Vuex; designed hybrid memory architecture with Redis and PostgreSQL; built a multimodal data interpreter integrating multiple LLM providers (OpenAI, Anthropic, Google). Implemented enterprise orchestration to manage concurrent LLM requests and tool calling, reducing research workflow time from hours to minutes.

Software Development Engineer at Microsoft

Optimized memory usage across cloud service infrastructure by 25% through efficient data structures and caching; saved $50K+ annually for 1M+ daily requests. Developed and maintained automated monitoring tests using C# and Azure DevOps, achieving 99.9% uptime and reducing mean time to detection from 30 minutes to 5 minutes. Implemented frontend features with TypeScript, React, improving user experience by 30% and reducing UI issues by 45%. Resolved 200+ support tickets via log analysis, maintaining 4.8/5 CSAT and reducing average resolution time from 48 to 12 hours.

Research Assistant & Full Stack Developer at Doxey Lab at University of Waterloo

Built a gene visualization web app (annoview) using Flask, MongoDB, MySQL, and Docker on Google Cloud; served 500+ researchers globally with 99.5% uptime. Optimized GTDB queries with indexing, achieving 70% performance gains and reducing analysis time from hours to minutes. Designed a novel genomic clustering algorithm combining hierarchical clustering with Smith-Waterman distance metrics, implemented via dynamic programming, yielding 25% accuracy improvement. Collaborated with an 8-person team using Agile; contributed to 3 peer-reviewed publications and presented findings at 2 international conferences.

Machine Learning Developer Intern at ByteDance AI Lab, TikTok Security Team

Developed large-scale fraud detection system using Hadoop/Spark processing 100M+ daily user interactions; achieved 92% accuracy and reduced false positives by 35%. Implemented LSTM-based sequence analysis and FP-Growth for behavioral pattern recognition; led to 15K+ fraudulent accounts detected and protected $2M+ in platform value. Built automated web crawler using Selenium monitoring 50+ underground markets, contributing to 200+ account takedowns and improved platform security.