LIANGJUN (LANCE) SONG, PHD

FORWARD DEPLOYED AI ENGINEER · ENTERPRISE LLM & AGENTIC SYSTEMS

FORWARD-DEPLOYED CLIENT ENGAGEMENTS

Ultra OT (occupational therapy practice)
2026
Lead Engineer — onsite client engagement
  • Engaged onsite with a practice that had no single view of its operations — case status, scheduling and billable time lived across spreadsheets, calendars, and email. Ran discovery with clinicians and admin staff to map the real workflow before committing to a build.
  • Scoped the smallest useful release first — case flow and Jira-style ticketing — then grew it into the practice's full internal operations system as sole engineer: calendar scheduling, client intake imports, and time/finance visibility (Next.js, TypeScript, Supabase).
  • Solved their hardest scheduling problem by clustering clients geospatially on driving distance via the Google Routes API, turning ad hoc visit planning into route-aware scheduling; iterated on it in place as the team used it.
Termly (healthcare)
2026
Lead Engineer — onsite client engagement, regulated healthcare
  • Delivered a document AI system for Australian medical contract automation, replacing slow and inconsistent manual contract review in a regulated healthcare setting.
  • Built the full pipeline as sole engineer: scanned PDF extraction, OCR correction, structured clause and entity extraction, Medicare/PBS identifier validation, and automated risk scoring (FastAPI/Pydantic, Claude Sonnet, Azure Document Intelligence, Tesseract, Docker).
  • Designed a provider-switchable LLM layer with token and cost tracking, plus domain validation rules and human review points, so output was checkable rather than trusted blindly.

WORK EXPERIENCE

WiseTech Global
Apr 2025 – Jun 2026
AI Engineer - AI/ML Group
  • Architected two end-to-end LLM systems for production support — CWBot, a support chatbot, and TriageAgent, an automated ticket-triage agent — combining LangGraph orchestration, FastAPI services, Azure OpenAI, enterprise APIs, and durable PostgreSQL-backed state.
  • Designed RAG workflows with dynamic question rewriting, custom retriever factories, Azure AI Search integration, and heterogeneous internal data sources to improve how users find and act on operational knowledge.
  • Implemented guardrails for PII filtering and response steering, plus checkpointing and recovery patterns for reliable multi-turn product experiences.
  • Instrumented the production layer around the model: request tracing, structured logging, and token/cost accounting to keep behavior and spend observable once deployed.
  • Led container-first CI/CD modernization for AI services (Docker dev containers, GitHub Actions, readiness checks, environment parity), with test coverage and release-readiness gates across backend, frontend, and system suites.
Redbubble
Mar 2023 – Jan 2025
Data Scientist - Search & Recommendation Team
  • Led search and recommendation enhancements with Marqo vector search and GCP Vertex AI MLOps pipelines, improving add-to-cart rate by 0.5% and CTR by 10%.
  • Owned production ML workflows end to end — offline evaluation, experiment design, deployment, and post-launch metric review — communicating results and model logic to product stakeholders.
  • Designed ML/data infrastructure processing 100M+ user events for feature extraction, search relevance analysis, and downstream serving workflows.
  • Drove GA4 analytics migration to ensure data reliability across internal dashboard-driven A/B testing frameworks.
Redbubble
Jan 2021 – Mar 2023
Data Scientist - Content & Discovery
  • Shipped production content-classification and moderation systems using language-image models, reducing intellectual-property moderation risk and improving operational throughput.
  • Built image duplicate-detection pipelines and data quality/anomaly-detection workflows for safer model and pipeline operation.
  • Ran content tagging, taxonomy, and SEO experiments to improve product discovery.
LIANGJUN (LANCE) SONG, PHD PAGE 2

SELECTED AI SYSTEMS

CallForMe (GitHub collaboration)
2026
Co-Creator & Lead Developer
  • Collaborating with a partner developer on an AI phone assistant MVP — call records, transcripts and summaries, AI risk analysis, and interactive hold-for-me call flows — built on Next.js/TypeScript with WebSocket real-time services.
Agentic Workbench (Remote Agent Workbench + AgentForge + BranchFlow · GitHub)
2026
Creator & Lead Developer
  • Built a multi-agent coding-agent collaboration system: Remote Agent Workbench (browser orchestration for Claude Code and Codex with live PTY terminal streaming, task persistence, and allowed-directory safety boundaries) and AgentForge (control plane running coding agents in parallel across isolated git worktrees, with PM2 management and auto-commit scheduling).
  • Built BranchFlow, a React/TypeScript nonlinear writing and prompt-orchestration workbench with branching story state, React Flow mind maps, and Obsidian Canvas export.
Voice AI Companions (RageZone · Elder Companion · GitHub)
2026
Creator & Prototype Builder
  • Built two voice-first AI companions: Elder Companion, an Expo voice app with ElevenLabs speech, a family dashboard, and a daily summary workflow; and RageZone, a WeChat Mini Program real-time communication assistant with Deepgram speech recognition, plus a corpus and evaluation scripts for conversation quality.

OPEN SOURCE

SGLang Framework (LMSYS / UC Berkeley)
Feb 2025 – Present
Open Source Contributor
  • Contributing runtime and serving optimizations to the SGLang LLM inference framework for state-of-the-art models including DeepSeek R1, Llama 3, and Qwen, spanning backend runtime, distributed serving, and prompt-language features.
  • Using AI coding agents (Claude Code) to navigate a large unfamiliar codebase, draft PRs and review diffs — a human-in-the-loop workflow for getting productive in someone else's system quickly.

TEACHING & STAKEHOLDER ENABLEMENT

Australia IT Group · ANZCC
Feb 2026 – Jun 2026
Guest Instructor & Workshop Facilitator
  • AI Engineer Bootcamp: delivered 12+ sessions on production RAG, vector search, multi-agent systems (LangGraph, AutoGen, CrewAI), QLoRA fine-tuning and RAGAS evaluation; designed Dispatch.AI, a 6-week project taking 13 students from scaffold to a deployed AI booking assistant.
  • ANZCC AI workshop: designed and delivered a hands-on workshop introducing agentic AI workflows to accounting professionals, adapting technical content for a non-technical business audience.

RESEARCH EXPERIENCE

Microsoft Research Asia
Jul 2012 – Jun 2013
Research Intern - Web Search and Data Mining Group
  • Developed Autosub, a collaborative project for generating video subtitles using advanced speech-to-text APIs.
  • Contributed to web page mining projects using data mining and time-series analysis techniques.

PUBLICATIONS

Towards Efficient Personalized Ranking
PhD Thesis, RMIT University
2020
Incremental Preference Adjustment: a Graph Theoretical Approach
The VLDB Journal (Core Rank A*)
2020
Continuous Summarization over Microblog Threads
DASFAA (Best Student Paper Award Runner-Up)
2017