Liam (Chengxuan) Han

• Built and shipped Ask Kimia with multi-step tool execution, streamed responses, failure handling, and run-level tracing using TypeScript, Node.js, and NestJS.
• Built a React and TypeScript review-and-apply workflow so users can inspect agent-proposed updates before changing workspace data.
• Built Access Data Scopes with NestJS and PostgreSQL, enforcing product and document permissions across the application, public APIs, and AI assistant retrieval.
• Designed database-backed organisation module access and entitlements, with administrative controls for customer modules, roles, and product tiers.
• Built event-driven notifications with BullMQ daily and weekly digests, per-user email limits, Resend idempotency, and PostHog engagement telemetry.
• Developed agent tools, run lifecycle, failure recovery, and server-side assistant capabilities across NestJS and Python FastAPI services.

Delivered Access Data Scopes from design to production within a month. Built a notification system supporting 30+ in-app and email topics, and an Ask Kimia workflow that brings knowledge base updates from lengthy SME-dependent waits to seconds.

• Built React and TypeScript interfaces for agent configuration and analytics, backed by Python, Flask, FastAPI, SQL Server, and Azure Cosmos DB.
• Built a configuration pipeline translating customer settings into runtime agent workflows, FAQs, voice, and booking rules through APIs and Azure Cosmos DB.
• Built an automated call analytics pipeline using Azure services and AI classification, turning conversations into searchable records and customer-facing insights.
• Built Python conversational agents combining LLM intent detection with multi-turn booking, change, cancellation, FAQ, and human-handoff workflows.

Replaced manual CSV configuration updates with self-service settings, reducing turnaround from days to seconds. The conversational AI systems handled 31K+ calls, 3K+ bookings, and 3K+ SMS events in a monthly production reporting period.

• Built delivery platform interfaces and dashboards using React, TypeScript, Axios, Next.js, Node.js, and GraphQL.
• Built GPT-powered workflows to generate articles for client websites and optimize SEO.
• Contributed to website UI improvements.

Automated article generation to reduce manual content creation and supported client website improvements through SEO tools and UI updates.

• Contributed to data handling, REST APIs, model training, and prediction workflows.
• Worked with structured datasets, file inputs, and client-server communication.
• Used Docker-based containerization to support application logic and model processing.

Enabled users to upload clinical data, select machine learning models, and view prediction results.