Built to finish the system.
Nixense Vixion is an engineering studio for difficult intelligent systems. We take on the intelligence and the software around it, from the first uncertain question to production operation.
Pakistan · Working internationally
A recognition model has to support a shelf decision. Retrieval has to return the shot a filmmaker meant. A recommendation has to survive an incomplete question and a changing catalogue.
Those are the kinds of problems we work on. They need judgment about data, evaluation, interfaces, cost, and failure, as well as the model itself.
Our team can build the web and frontend interfaces, mobile applications, data pipelines, cloud infrastructure, and operational tooling needed to put the intelligence to use. Those disciplines belong to the same engineering responsibility.
See the systems we have built ↗
Ahmer Saleem
Ahmer’s engineering record spans shelf recognition, cinematic search, conversational product discovery, controlled editing, and privacy-aware research.
It includes five years of AI architecture and model lifecycle work for Shelfr, multimodal retrieval for ShotDeck, and leading architecture and a five-engineer team for Biscuit AI.
The case studies describe the contribution and evidence for each engagement. The Edwards Lifesciences collaboration is presented separately as research.
We define the intended behavior, examine the available evidence, and agree what the system must do before calling it ready. The work includes implementation and evaluation, alongside the software and infrastructure required to run it.
Decisions, limitations, and operating responsibilities need to be understandable to the people who will use and maintain the system.
How we approach the work ↗