Hello, my name is
Mattheus
Lim
Statistician turned AI engineer
AI/ML engineer specialising in multi-agent systems and applied computer vision, most recently at Jellyfish leading the AI layer of a brand compliance platform for a Tier-1 UK bank. Melbourne-based and open to AI/ML engineering roles.
About
I'm an AI/ML engineer with a statistics background, specialising in multi-agent systems and applied computer vision. Most of my work sits where the modelling meets the client: scoping what an agentic system should actually do, building it, then explaining it to the people who have to trust its output.
Before moving into AI engineering I spent five years in data science and analytics across e-commerce, health tech and performance marketing, which is where I picked up the parts that aren't modelling. I've mentored data scientists on experiment methodology, run engagements with enterprise stakeholders across North American and European markets, and managed a data team through onboarding. The hardest problem in most projects is still agreeing on what "correct" means before anyone builds anything.
Coding Languages
AI & Machine Learning
Tools & Platforms
Ways of Working
Experience
2 roles over Oct 2022 - Jul 2026
Visit the Jellyfish websiteJellyfish is a global performance marketing agency
Senior AI/ML Engineer
Nov 2024 - Jul 2026 · London, UK → Melbourne, VIC (transferred Feb 2026)
- Designed and led the AI layer of a multi-agent brand compliance API for a Tier-1 UK bank, now in client UAT ahead of a planned 30,000-user internal rollout
- Orchestrated a snapshot and routing agent feeding parallel image/video analyst agents and a summary reporter via CrewAI, evaluating creative assets against 400 production QC rules within a negotiated 15-minute SLA
- Generated QC rules from mixed-layout PDFs via deterministic topic-based chunk iteration with an MLLM, after a Qdrant-backed RAG retriever produced duplicated and incomplete rules on high-level queries
- Applied transfer learning to 2 computer vision tools backing the agents' deterministic checks: YOLOS (TensorFlow) for logo presence detection and ResNet (PyTorch) for font classification, trained on thousands of labelled examples
- Accelerated report generation from days to minutes by developing Jellyfish Social Agents™, a CrewAI multi-agent system now used by ~25 marketing intelligence analysts to audit any Instagram, TikTok or YouTube handle
- Built an internal MCP-powered agent on Google Agentspace for chat-based qualitative and quantitative analysis of those reports, and an automated GCP pipeline to fine-tune OpenAI models for fashion product descriptions, halving deployment time
Senior Data Scientist
Oct 2022 - Oct 2024 · London, UK
- Advised 4 enterprise clients across NA and EU markets on incrementality measurement, standardising GeoLift experiment tooling in Python and mentoring 5 data scientists on the methodology
- Delivered Data-Driven Attribution engagements under a Google Partnership, productionising GA4, GCP, SQL and Airflow pipelines for automated user transaction insights
- Secured 3 client engagements under the same Google Partnership, auditing customer lifetime value (CLV) data and recommending predictive modelling enhancements to bidding strategies
Education
Bayes Business School
Visit the Bayes Business School websiteMSc Business Analytics / Data Science (Distinction)
- Applied Research Project with Rolls-Royce to optimise existing speech assistive software for MND patients utilising Stanza NLP pipelines and pre-trained dialog models from Facebook (Meta) ParlAI Python framework
- Ranked #1 in Network Analytics Project on “Regulatory Impact on Ethereum Ecosystem”; selected out of 60 students to contribute to a research paper about “Role of Regulation on Cryptocurrency Markets”
- Machine Learning, Deep Learning, Data Visualisation, Applied NLP, Data Management Systems (Python, R, SQL)
University College London
Visit the University College London websiteBSc Statistics
- First Class Honours dissertation – “Improving the Estimation of the Weight of Drugs Impregnated in Clothing
Projects
Customer Segmentation through Machine Learning
Customer segmentation using K-Means clustering on customer dataset in R.
Insurance Claim Fraud Detection
Autoencoder anomaly detection of fraudulent insurance claims using TensorFlow.
Time Series prediction with RNN and CNN
Developed Recurrent and Convolutional Neural Networks in TensorFlow to predict the operating mode of a wind turbine based on 2 time series data from sensors.
Impact of China cryptocurrency ban on Ethereum ecosystem
Analysed Decentralised Application (DApp) and whales' activities through exploring the network of Ethereum transactions before and after the China crypto ban.
Elon Musk's influence on DogeCoin
Identified whether Elon Musk's DogeCoin tweets influences the cryptocurrency's price.
Effect of Covid-19 on UK business foundation
Elucidated the impact of the Covid-19 pandemic on UK firms and sectors through EDA and visualizations.
Machine Learning Shiny App
An interactive web app that predicts the winner of a League of Legends match through Decision Tree and Random Forest models.
What makes TikTok videos popular?
Demystified TikTok's popularity through EDA and visualisations.