Haven Feng

– Led development of internal AI assistant for support and operations teams using OpenAI APIs, LangChain, and LangGraph to answer queries on system documentation, logs, and production metrics. – Implemented RAG (Retrieval-Augmented Generation) pipelines using FAISS and AWS Bedrock / OpenSearch as vector stores to enable natural language access to platform configuration data and troubleshooting guides. – Integrated AI workflows with internal APIs and AWS services to automate FAQ responses, data summarization, and engineering knowledge retrieval, improving issue resolution time by over 30%. – Experimented with fine-tuning prompt chains and tool-calling agents to extend assistant capabilities, enabling contextual actions such as metrics queries and deployment checks. – Architected and maintained a modular, large-scale backend system for a multi-tenant gaming platform, leveraging Node.js and AWS Lambda behind Amazon API Gateway to deliver high-availability, low-latency services for payment, bonus, wallet, and third-party integrations. – Developed scalable, feature-rich front-end applications using React.js, TypeScript, and Redux, delivering seamless user experiences for multi-tenant gaming dashboards including real-time metrics, player engagement analytics, and bonus configurations. – Optimized front-end performance by implementing server-side rendering (SSR) with Next.js and lazy loading, reducing initial page load times and enhancing core web vitals for a responsive, enterprise-grade platform. – Designed and maintained robust cloud infrastructure using Infrastructure as Code (IaC) principles, leveraging AWS CloudFormation templates to automate provisioning of VPCs, databases, S3 buckets, DynamoDB tables, and CloudFront distributions. – Built event-driven workflows integrating Apache Kafka for asynchronous processing and reliable inter-service messaging. – Authored contract-first REST APIs with OpenAPI/Swagger, enabling consistent request/response validation and partner integration. – Modularized shared logic with AWS Lambda Layers (e.g., validation/payment layers), improving reuse and reducing bundle sizes across microservices. – Implemented payment provider integrations at scale (dozens of adapters), handling webhooks, signing/HMAC, and idempotent callbacks in Node.js. – Standardized AWS SDK usage with compatibility layers, streamlining migrations and interop across SDK versions. – Automated deployments and environment tasks with shell scripting, integrating into Git-based CI/CD pipelines. – Instrumented services with structured logging and metrics, enabling operational visibility and rapid troubleshooting across domains (auth, bonus, risk, payments).

– Designed and implemented a comprehensive pharmacovigilance platform processing over 10,000+ scientific articles monthly using python microservices, AWS Lambda, and DynamoDB. – Built multi-tenant literature processing system handling 50+ pharmaceutical clients simultaneously with automated ETL pipelines, LLM integration, and real-time data validation. – Developed RESTful APIs using FastAPI for pharmaceutical literature ingestion and drug safety monitoring systems, implementing async endpoints for high-throughput document processing, real-time data validation, and multi-tenant authentication with automatic OpenAPI documentation generation. – Developed cloud-native infrastructure using Terraform IaC across 15+ AWS services including lambda, S3, DynamoDB, API gateway, and OpenSearch. – Implemented AI-powered literature extraction engine integrating OpenAI GPT, Google Gemini and AWS Bedrock for intelligent query optimization and content processing. – Built custom NLP pipelines using Transformers library for drug safety signal detection, processing 100GB+ of scientific literature with 99.5 uptime. – Designed and implemented an intelligent query builder using Python and SQL, integrated with redundancy removal and optimization algorithms, reducing manual query construction time by 70%. – Developed response web applications using Nuxt.js/Vue.js for drug safety monitoring dashboards and literature management interfaces, serving 500+ concurrent users. – Built real-time data visualization and reporting tools using D3.js, Plotly.js, and Chart.js, streamlining pharmaceutical safety analytics and enabling faster decision-making while meeting regulatory compliance requirements. – Engineered multi-tenant user interfaces with role-based access control (RBAC) and personalized dashboards using frameworks like Vuetify and Tailwind CSS, tailored for distinct pharmaceutical clients. – Orchestrated CI/CD pipelines using GitHub Actions for automated testing, building, and deployment across multiple environments(dev/test/prod). – Managed containerized applications using Docker and AWS ECR, implementing blue-green deployments and automated rollback capabilities. – Implemented comprehensive monitoring and observability using CloudWatch and ELK, custom logging frameworks, and performance metrics tracking across distributed systems. – Developed comprehensive test suites using Jest, Mocha, Chai for unit tests, integration tests, and performance benchmarks, achieving 90%+ code coverage.

– Built and managed cloud infrastructure on AWS using Terraform, containerizing services with Docker and Kubernetes, while leveraging AWS Lambda for event-driven workloads. – Integrated MapBox to enable the visualization of real-estate trends, including property price heatmaps, neighborhood clustering, and GIS-powered map components. – Designed intuitive features like dynamic clustering of nearby properties for improved performance at various zoom levels and interactive tools (pop-ups with property details, boundary-drawing for filters) using React and Next.js. – Leveraged Apache Kafka for ingesting and processing large datasets in real-time, enabling smooth rendering of heatmaps and fast response times for dynamic property queries. – Used Amazon ElastiCache to cache frequently accessed data such as clustering, map data and aggregated data for better performance. – Deployed backend logic using AWS Lambda, optimizing performance for millions of concurrent users with techniques like provisioned concurrency to reduce cold-start latency. – Configured a RDS Proxy between Lambda functions and a PostgreSQL database hosted in VPC for managing connection pooling, effectively solving connection bottlenecks. – Pre-aggregated PostgreSQL data using event-driven pipeline, enhancing user experience by accelerating UI responsiveness. – Designed an OpenAI-powered chatbot that provided property purchasing insights and investment recommendations, using domain-specific training to ensure high-quality answers.

– Developed and deployed a data-driven alert system to monitor and analyze the public sentiment towards refugees using social media data from platforms like Twitter, Linkedin, Reddit, and Facebook. – Migrated backend logic from Next.js to AWS Lambda functions, ensuring a secure serverless architecture within a VPC for a PostgreSQL database, enabling seamless and private connectivity after tightening database security policies. – Implemented real-time visualizations for sentiment data using React, Next.js, and Chart.js to provide interactive dashboards, showcasing trends like timeline sentiment analysis, keyword insights, and regional distributions of discussions related to refugees. – Established CI/CD pipelines and infrastructure using AWS Amplify, Jira for agile project management, and IaC (CloudFormation) for scalable deployments. – Enhanced scalability and reduced latency by implementing an ETL pipeline using AWS Lambda, S3, RDS (Aurora), and DynamoDB for robust and automated data handling.