01 / Selected work
Engineering with impact.
Selected initiatives from my professional experience, spanning enterprise AI, data platforms, and distributed systems.
01 / AI Advance Consulting
Multi-agent project assistant
Designed a modular LLM-powered assistant that connects meeting conversations to project execution.
40%reduction in manual coordination
GPT-4oLangGraphRAGMCPAzure AI Search
Explore the implementation
- Automated meeting summaries, task extraction, and Jira/Slack updates.
- Built custom Model Context Protocol servers for standardized information exchange between agents.
- Used Power BI to monitor accuracy, response times, and engagement; deployed through Azure OpenAI and Azure ML.
02 / AI Advance Consulting
Retail sentiment intelligence
Built a sentiment engine to analyze consumer reviews, support transcripts, and retail channel feedback at scale.
92%accuracy across 50,000+ daily feedback records
BERTPandasAirflowRedisDocker
Explore the implementation
- Fine-tuned BERT models and performed exploratory analysis to identify satisfaction drivers and purchasing trends.
- Automated ingestion and retraining with Airflow and Docker; added Redis caching to reduce inference latency.
- Insights contributed to a 15% increase in customer satisfaction and brand loyalty through messaging and product improvements.
03 / KREBSoft
Multilingual document intelligence
Engineered resilient processing for Japanese, English, and Chinese Safety Data Sheets (SDSs).
95%document classification accuracy
Document AIClaudeGPT-4GeminiPython
Explore the implementation
- Automated document type, language composition, and format quality detection.
- Implemented fallback mechanisms between LLMs to improve processing resilience.
- Translated complex regulatory requirements into technical specifications with chemical domain experts.
04 / KREBSoft
Chemical data platform
Built a pipeline that collects chemical substance data from regulatory sources into a centralized information system.
24-hourscheduled crawler deployment cycle
BigQueryCloud RundbtGitHub Actions
Explore the implementation
- Implemented parallel-processing crawlers for NITE CHRIP and TSCA regulatory databases.
- Built cloud-native ETL with Cloud Storage and BigQuery; established dbt staging models and source configurations.
- Added dbt_expectations quality checks and automated crawler deployment as Cloud Run jobs.
05 / New Dawn AI
Distributed backend services
Developed Node.js and Go services for transaction processing, authentication, validation, and risk assessment.
5Go microservices targeted for performance improvements
GoNode.jsgRPCKafkaGraphQL
Explore the implementation
- Optimized HTTP clients, addressed memory leaks, and implemented context-based timeouts.
- Improved gRPC connection management and DNS caching behavior in Kubernetes for resilient rolling updates.
- Migrated REST APIs to GraphQL on a Go backend and integrated with a React frontend.
06 / New Dawn AI
Cloud delivery & application security
Connected secure APIs, asynchronous messaging, and automated delivery for scalable applications.
GKEcontainer orchestration and delivery
KubernetesDockerPub/SubOAuth 2.0JWT
Explore the implementation
- Built CI/CD pipelines with GitHub Actions, Docker, and Google Kubernetes Engine.
- Used Google Cloud Pub/Sub for reliable asynchronous messaging in healthcare application microservices.
- Implemented OAuth 2.0 and JWT authentication with Passport.js; integrated Elasticsearch for improved data retrieval.