AI Developer Lab
Projects
PriceEcho
Groceries, fuel & local prices
PriceEcho app
AI-Powered Retail Intelligence
An AI-driven platform for collecting, normalising, and analysing grocery, fuel, and local pricing data. PriceEcho combines machine learning, geospatial search, and intelligent product matching to deliver structured retail intelligence across multiple markets.
ATTENTION-TO-SSM
Attention-to-SSM distillation research
This project explores methods at the structural intersection of attention-based Transformers and linear State Space Models - specifically, linearizing attention layers into SSM (Mamba-2 / Gated DeltaNet) layers via distillation, to study how much of multi-head attention's associative recall can be preserved under linear context-length scaling, at small (sub-2B) scale and for domain-specialized tasks.
Zews
Research & Development
Exploring new foundations for machine intelligence. Experimental research into computational architectures, training methods and datasets spanning mathematics, geometry, dynamical systems and physical modelling.
Price Echo AI
A retail price-lookup assistant built on a custom language model
Most price tools either scrape rigidly -breaking the moment a retailer changes their markup - or throw the whole page at a large model and hope. Price Echo AI does neither. It uses a purpose-built hybrid SLM architecture, trained from scratch on consumer hardware, paired with a deterministic parsing layer that handles what code does better than language models do.
The result answers questions about real product pages: what something costs, what it costs per kilogram, what it cost before the discount, what three of them would come to. It says so when a product isn't on the page rather than inventing a plausible price - a failure mode that makes most naive approaches unusable for anything commercial.
Currently tested across major grocers. In development.
