FIG. 01
in study
AI systems that survive production
Patterns for shipping LLM systems that hold up under real load: multi-agent orchestration, incident response, cost, and sandboxing the agents themselves.
aLLMbAWScDocker
no entries yet
ML · Products · Infrastructure
1% better,
every build.
I'm Asif and welcome to my attempt at keeping my work and thoughts organized. Below you'll find the problems I've worked on, what they taught me, and the bits worth sharing. Start anywhere, or flip the lens to the tech behind it.
Writing01
Notes worth sharing
Lens
Read as↳practical, how to ship it
FIG. 02
in study
Teaching machines to read the earth
Productionising ML on 3D volumetric data that maps the subsurface: representation, training at volume, and getting models off the notebook and into a pipeline.
aMLb3DcPython
no entries yet
FIG. 03
in study
An AI life coach, built from scratch
The Experiment: custom memory, open-source models, and a real evaluation framework. The longest-running build, and the one throwing off the most write-ups.
aLLMbEval
no entries yet
FIG. 08
in study
Shipping & operating products solo
Lessons from building and operating SortedOut and Rinqer end to end: turning a gnarly rules engine into something strangers trust, and what traction really costs.
aCloudbPython
no entries yet
FIG. 04
to examine
LLM-augmented data pipelines
Where language models genuinely earn their place in a pipeline: extraction, routing, and validation, and where a plain function is still the right call.
aLLMbPythoncCloud
no entries yet
FIG. 05
to examine
Clustering: finding the shape of data
Unsupervised and supervised clustering in practice: when each earns its keep, evaluating the result, and not fooling yourself with pretty plots.
aMLbPython
no entries yet
FIG. 06
to examine
Cloud platforms & data landing zones
Platform engineering for data: landing zones, guardrails, and the boring infrastructure that makes everything above it possible.
aCloudbAWScPlatform
no entries yet
FIG. 07
to examine
Buffett-style fundamental analysis, in code
Systems that do fundamental analysis with discipline: encoding the questions a value investor asks, and keeping the model honest.
aMLbLLMcPython
no entries yet
Side projects02
The ones I chip away at after hours
Get in touch
Notes from the next
hard thing, now and then.
No schedule, no funnel. Just a short note when a build teaches me something worth passing on.