About
Hi, I'm Leo
I maintain Forcebench. I'm an AI engineer and software consultant, and I run Azul Labs, the company behind it, from the Sunshine Coast in Queensland.
I've been building software for 15 years: AWS, Salesforce, Python and, more recently, a lot of AI work with Claude and other LLMs. On the Salesforce side that's Apex, Lightning, CI/CD and deployment pipelines.
Lately a lot of my work is setting up and fine-tuning local models for teams that need to keep data in-house, and writing evals for agents before they go live. About a quarter of my own work is handled by Claude and the rest by local models. When someone asks me which model is better, my answer is always the same: let's test it.
Why Forcebench
Forcebench is where those two halves meet. Salesforce teams are being told AI will write their Apex, build their Flows and answer their admins' questions. Some of that is true. Much of it is untested. The public benchmarks are mostly Python and puzzles; they can't tell you whether a model knows what a mixed DML error is or will bulkify a trigger without being asked.
So I'm building the benchmark I wanted: real tasks, graded by running them, with error bars on every number, in the open. Final answers, grades and token counts for public tasks are published in the repo; reasoning traces are available on request through the gated Hugging Face dataset; nothing from the private pool is published except approved aggregate scores. If AI isn't the right tool for a job, I'd rather the numbers say so.
Azul Labs
Azul Labs (Azul Labs Pty Ltd) is my company. It runs Forcebench, and runs it privately for two kinds of Salesforce team:
- Choosing a model to run yourself? A private benchmark of your shortlist: open-weight models on infrastructure you control, tested on your kind of Salesforce work, with a written recommendation. That's sovereign AI.
- Building or buying a Salesforce AI tool? A private evaluation of your tool on held-out tasks it has never seen, against API baselines and Local ones (served on Forcebench’s own hardware), with a written report on where it holds up and where it fails.
Both are a fixed fee, quoted up front; Private evaluations has how they work. Beyond Forcebench, Azul Labs also helps startups and scale-ups with agents, local models and evals.
I also keep a digital garden of perpetually incomplete notes on AI, neuroscience, maths and whatever else I'm learning.
Get in touch
The easiest way to reach me is leo@azl.au, or through azl.au. I'm also on LinkedIn, GitHub and X.
Private evaluations
Want this for your team?
Azul Labs runs Forcebench privately for two kinds of Salesforce team.
Choosing a model to run yourself?
A private benchmark of your shortlist
Open-weight models at the quantisations and reasoning efforts you could run on infrastructure you control, tested on your kind of Salesforce work, with a written recommendation.
- Private tasks written from your work only ever run on local models; they are never sent to an API.
- Re-testing after any change of model, quantisation or engine is available.
- Typical turnaround is 2–3 weeks; large models take about a day of machine time per configuration.
Building or buying a Salesforce AI tool?
A private evaluation of your tool
Your tool on held-out tasks it has never seen, against API baselines (models reached through a hosted service) and Local ones (served on Forcebench’s own hardware), with a written report on where it holds up and where it fails.
- It can start as a smaller pilot, about 30 held-out tasks with your tool against two baselines, before a full evaluation.
- Repeat evaluations after your releases are available.
- Typical turnaround is 2–3 weeks.
Fixed fee, quoted up front. The quote depends on the number of tasks and the number of models, tools or configurations tested.
Or email leo@azl.au · azl.au