Fisher Kuan

• Lead end-to-end GenAI engagements: discovery, architecture, hands-on implementation, and go-live support.
• Championed company-wide agentic engineering adoption: led research and industry networking, hosted internal workshops on best practices, and built an internal scaffolding repository to standardize and accelerate team-wide adoption.

LaunchPad — Built a modular RAG accelerator framework on Azure for enterprise chatbot deployments; extended it with an agentic architecture to automate multi-step workflows. Tech: Python, FastAPI, Azure OpenAI, LangChain, Vue 3.

Fiducia 4.0 (Puratos) — Developed ML system predicting QC scores across multi-plant food production. Integrated raw material, process, and QC data sources into automated Azure ML pipelines. Enabled proactive quality control for Belgian and Mexican facilities.

• Designed and implemented GenAI solutions (document intelligence, process optimization) using LlamaIndex and LLM APIs.
• Mentored engineers in applied GenAI patterns, delivery practices, and client communication.

BioSignature 2.0 (Johnson & Johnson) — Led consolidation of multiple microservices into a streamlined event-driven ML pipeline, reducing prediction turnaround from days to hours. Tech: Python, AWS (Step Functions, SageMaker), KServe, Argo, Kubernetes, Jenkins.

• Coached junior ML engineers across the full lifecycle with emphasis on engineering quality.

Dynamic Pricing (Proximus) — Built RL-based dynamic pricing system driving 100% profit growth in nine months. Tech: Python, PyTorch.

Chatbot (BNP Paribas Fortis) — Developed a secure on-premises chatbot service using Llama 2.

• Streamlined pipelines with Azure Data Factory/Synapse and Azure DevOps.

Demand Forecasting (Unilever) — Developed sale and demand forecasting applications to reduce shortages and overproduction, shaping production strategies across facilities. Tech: Databricks, PySpark, Azure (Azure ML, Cognitive Services).

• Delivered NLP, web scraping, and optimization projects end-to-end; built feature extraction pipelines with BigQuery.

• Developed ML models for semiconductor packaging optimization; reduced product development timelines by 20%.
• Led thermal/structural analysis workstreams influencing reliability and design decisions.