About Me

I am an applied ML researcher and systems engineer specialising in information retrieval, recommendation, local AI and data-intensive software.

My work usually begins with an ambiguous or unusual problem that does not fit a standard product. I turn those requirements into experiments, system designs and deployable software, working across modelling, architecture, implementation, evaluation and delivery.

I am available for contract and consulting engagements, from focused technical investigations and prototypes through complete specialist systems.

Discuss a project - LinkedIn - GitHub

What I Can Help With

  • search, retrieval, ranking, reranking and evidence-linking systems
  • local and privately hosted LLM applications
  • RAG, cited synthesis, structured generation and agent workflows
  • recommendation, personalisation and online-learning systems
  • document ingestion, OCR, indexing and monitoring pipelines
  • model and retrieval evaluation, benchmarking and performance analysis
  • ML platform architecture and production serving
  • technical discovery, prototyping and end-to-end product development

Selected Systems

Local Cited Research

I designed and built PrivateCite, a commercial Windows application for citation-backed search, multi-document synthesis and structured reporting.

The work includes a retrieval engine based on SPLADE sparse embeddings, sparse-array storage and Roaring bitmap indexes; document conversion and OCR; incremental filesystem monitoring; passage-level citations and highlighting; local inference across NVIDIA and AMD consumer GPUs; and an agentic analysis and report-generation workflow.

Real-Time Recommendation

I architected a production retail recommendation platform using multimodal product embeddings built from images, descriptions, categories and catalogue metadata.

The system combined Thompson sampling with recent sales behaviour and emerging product performance. Its low-cost in-memory CPU serving layer was designed for containerised autoscaling and real-time use. Deployments included major fashion and furniture retailers across the Asia-Pacific region.

Search and Personalisation

I architected a real-time news discovery, search and personalisation platform built on large-scale RSS ingestion, Cassandra and Elasticsearch.

The platform combined topic models, hashed interest profiles, MinHash similarity and lexical retrieval so it could personalise established topics while still handling new entities and terms outside the trained model.

High-Dimensional Search and Navigation

I designed methods for navigating large music, image and video collections using audio and visual descriptors, interpolation between examples and iterative PCA-based drill-down.

This work contributed to two granted US patents covering high-dimensional database search using textual and vector representations, and search using multiple examples.

Quantitative and Decision Modelling

Before specialising in ML systems, I worked as a senior econometrician and quantitative-modelling consultant.

Projects included applied econometrics, operations research, regional input-output modelling, infrastructure decision systems, Monte Carlo portfolio analysis and retail-development impact assessment.

Current Research

Current research includes persistent agents, external memory, structured world state, grammar-constrained generation and simulation systems.

SLIP is an experiment in programming-language design influenced by Lisp and Rebol, exploring modern syntax that maps closely to its abstract syntax tree, multidispatch, path-based computation and explicit state ownership. A persistent interactive world provides a demanding test case for ideas intended to generalise across domains.

Technical Range

  • Research and modelling: experimental design, benchmarking, econometrics, statistical modelling and online learning
  • ML and numerical computing: PyTorch, NumPy, SciPy, scikit-learn, statsmodels and Numba
  • Retrieval and NLP: SPLADE, sparse and dense retrieval, ranking and reranking, Roaring bitmaps, FAISS and Elasticsearch
  • LLM and agent systems: RAG, citation grounding, persistent agents, external memory and structured world state
  • Structured generation: Pydantic, JSON Schema, GBNF, regular-expression and PEG grammars, and Outlines
  • Inference and deployment: ONNX, ONNX Runtime, llama.cpp, ExLlamaV2, DirectML, CUDA, Vulkan, Linux and Windows
  • Data systems: PostgreSQL, SQLite, Redis, MySQL and MongoDB

Education

  • Master of Science in Computer Science (Machine Learning), Georgia Institute of Technology - GPA 4.0/4.0
  • Graduate Diploma of Commerce (Economics), University of Melbourne
  • Bachelor of Science (Computer Science), University of Melbourne

Selected Patents - Co-inventor

  • USPTO 8,832,134: Method, System, and Controller for Searching a Database Containing Data Items
  • USPTO 8,775,417: Method, System, and Controller for Searching a Database
  • USPTO 8,473,368: Method, System, and Controller for Providing Goods and/or Services to Consumers

Engagements

Available for contract and consulting work, including technical discovery, prototypes, architecture, specialist implementation and complete applied-ML systems.

Email tech@seerware.com