LUCALOWNDES

3rd of 49 university teams across the Asia-Pacific region, presented on stage at SupercomputingAsia 2026 in Osaka

Lead AI Researcher: DeepSeek-R1 671B (37B active) on H100 (NSCC ASPIRE-2A+) and two 8× NVIDIA H200 nodes

Fewer GPUs arranged deliberately beat more GPUs arranged by default: a tuned single node (PP2 · TP4 · DP4 attention) hit 17,417 tok/s, 81% over the 16-GPU baseline and 2.05× the per-GPU efficiency

CUDA version alignment alone was worth +35%. 40+ hand-tuned NCCL configurations never beat the defaults

HPC Track: optimised NWChem Fortran chemistry simulations across multi-node CPU clusters

Team: Josh Riantoputra (Captain), Nathan Culshaw (HPC Lead), Isaac Barnes, Giacomo Bonomi; mentor Simon Michnowicz

The work became the Firmus and HPC-AI Advisory Council partnership on SGLang best practices, and the ICPP 2026 paper

1st of 850+ participants and 135+ projects, youngest winner on record

Growth Garden, AI-powered personal development app

Architected the conversational AI system in Langflow with DataStax, part of the frontend, and the winning pitch

Accepted at Benchmarking in the Data Center (BID ’26), a workshop of the 55th International Conference on Parallel Processing

Shows that keeping collectives on NVLink and replicating within a node beats sharding across InfiniBand: 17,417 tok/s, 81% over the 16-GPU baseline

In a session alongside speakers from NVIDIA and the HPC-AI Advisory Council

Panellist with Jeff Adie (NVIDIA), Pengzhi Zhu (Director, HPC-AI Lab, HPC-AI Advisory Council) and Sandeep (IIT Palakkad)

Moderated by Dr Samar Aseeri and Dr Jens Domke

Joined as the first and only engineer. Designed and built Eve, an AI concierge that handles guest conversations end to end for short-term rental hosts, property managers and hotels, plus the property management system integrations and AI service architecture underneath it

The platform I built was the product investors backed: EvryStay raised A$4.5M at a A$17M post-money valuation, opening and closing the round inside a week

Hired and now lead five full-time engineers in person. I own the architecture, the AI service boundary, code review and the final technical call, while still shipping daily: 300+ merged PRs in my first seven months

Built Dummy, our Slack-native company agent that turns a feature idea into branches, a full-stack preview and draft PRs, so feedback happens on something concrete and the idea to shipped feature cycle is much shorter

Contributed optimisations to SGLang, the inference framework behind hundreds of thousands of GPUs worldwide, credited in the v0.5.5 release notes

Wrote deployment guidance for running SGLang on H200 clusters as part of a joint effort between Firmus, the HPC-AI Advisory Council, Monash eResearch and Monash DeepNeuron

Maximised inference throughput for DeepSeek-R1 (671B parameters, 37B active) across two 8× H200 nodes, on infrastructure from NSCC Singapore, NCI Australia and Firmus

Fewer GPUs arranged deliberately beat more GPUs arranged by default: a tuned single node (PP2 · TP4 · DP4 attention) hit 17,417 tok/s at 2.05× the per-GPU efficiency of 16 GPUs

The boring stuff mattered most: CUDA version alignment alone was worth +35%, and 40+ hand-tuned NCCL configs never beat the defaults

3rd of 49 university teams, presented on stage at SupercomputingAsia 2026 in Osaka. The work became the Firmus partnership and the ICPP 2026 paper

Implemented Neural Cellular Automata methods from recent papers in PyTorch to model self-organising systems such as embryonic gastrulation, plus on-device and cost-function optimisations using WebGL

Ran hands-on workshops in Victorian high schools on how deep learning models work and how to use generative AI well

Melbourne, VIC, best role to date

Built the Scag Mechanic and IM Advisor assistants for model comparison, troubleshooting and service recommendations

Structured proprietary product data, shipped embeddable chat widgets for web and mobile, and put brand persona and legal guardrails around both