Jiajun Lim
AI engineer building production LLM systems across frontend, backend, and infrastructure.
I picked AI as my major before ChatGPT made it the obvious choice, mostly on a hunch that it would end up mattering. That bet turned into building production systems: most recently a multi-agent chatbot that answers customer questions for an investment platform and a digital bank from one core engine, plus the backend and infrastructure that keep it running.
Stella
Multi-agent chatbot for financial services
One core engine answers customer questions across an investment platform and a digital bank. It handles recurring questions with automatic response checks and a handoff to a human when it cannot give a safe answer.
My contribution
Built across the agent orchestration, retrieval, streaming, and the backend and infrastructure that keep the system running.
Engineering highlight
Reworked the streaming pipeline after heavy concurrent load caused chat responses to stall for over a minute. Under the same load, responses now start almost immediately.
Technical details
Orchestrated with LangGraph. The system searches an internal knowledge base on Azure AI Search and Google Vertex AI, falls back to live web search when needed, streams responses in real time, and traces runs end-to-end in Langfuse. Responses go through automatic checks with a revision step before reaching the customer. Personal data is detected and redacted before processing.
Fintech lending integration
Backend and self-managed cloud infrastructure
On a client engagement, I owned the backend connecting a lending partner to a fintech platform, including loan and disbursement notifications and encrypted document exchange.
My contribution
Built a FastAPI service for authenticated partner callbacks and encrypted document exchange.
Cloud migration
Moved the full application, database, and uploaded documents from an office PC to DigitalOcean VMs. Provisioned and ran the cloud deployment while preserving the existing partner integration.
CVM stage classifier
Final-year research project, developed into CVM Studio
My final-year project explored six-stage cervical vertebral maturation classification using the CVM-900 research dataset. I later rebuilt the original Streamlit interface in Vue 3 with a FastAPI backend, so users can select a model and inspect its prediction alongside Grad-CAM heatmaps and class scores.
My contribution
Worked with ConvNeXt Small, DenseNet121, MobileNetV2, and EfficientNet-B1 for six-stage classification, with model-specific preprocessing and Grad-CAM visualization.
Application engineering
Built the Vue/FastAPI inspection workflow with original, heatmap, and adjustable overlay views. The backend validates uploads and keeps one model cached with one analysis at a time to bound memory use.
Comfortable across managed and self-managed infrastructure
Azure and GCP are my day-to-day at work. DigitalOcean VMs are what I run myself for a side project. I've also picked up Kubernetes, though it hasn't seen much production use yet. I switch between AI coding tools often, Claude Code and Codex among them, and have one model judge another's work before it ships. I like closing the agentic loop myself, automating the path from code to shipped, CI/CD included.
Outside of work, I follow macro and policy, and I like puzzling through options and equities. I'm drawn to decentralization for a similar reason: value should move as easily for someone without banking infrastructure as for anyone else. Years of K-dramas and K-pop got me to basic conversational Korean. I've recently picked up video editing, and filter coffee is a daily ritual.