HujraCart
Launched · in useOur first product, now in daily use by shopkeepers, field salesmen and customers. Sales, inventory and shop discovery on one map, on Android and Windows. See HujraCart →
Products, research and client builds. Each one is tagged with where it stands today, from prototype to launch.
Our first product, now in daily use by shopkeepers, field salesmen and customers. Sales, inventory and shop discovery on one map, on Android and Windows. See HujraCart →
Finds, reads and prices public tenders around the clock. A person only gives the final go-ahead.
A unique QR code and scratch-off PIN on every printed copy. Readers scan to check their copy is genuine.
Flags intracranial haemorrhage on CT scans by combining image features with patient data.
Operate a phone with your eyes alone. Iris tracking from the camera, paired with custom hardware, for people who can't use touch.
A digital-twin model that predicts when a machine will fail, using physics-informed features and a stacked ensemble, so maintenance happens before the breakdown.
HujraCart connects shopkeepers, field salesmen and customers. Launched and in daily use, on Android phones and Windows PCs.
FormPilot fills Google Forms and any website form from your Excel file, row by row. Launched and in use by customers. No external AI services, and your data stays on your own PC.
Pay with Easypaisa or JazzCash. You pay only for forms that are submitted successfully.
Bring us one piece or the whole thing. We handle the model and everything built around it.
Reading drawings, scans, photos and medical images, and turning pixels into data you can measure.
U-Net · YOLO · OpenCV · MediaPipe
Models for risk, failure and demand, tested honestly on data they have never seen.
XGBoost · ensembles · physics-informed
Agents that read documents, fill forms and run multi-step work, with a person approving what matters.
LLM APIs · RAG · tool use
Mobile and desktop apps, dashboards and backends that small teams can run without a server team.
Flutter · Python · Supabase · Cloudflare
Short loops, so you see measured results before anything expensive gets built.
01 · WEEK 1
We sit with the people who have the problem and collect real samples of their data.
02 · WEEKS 2–4
A prototype on your real data, with measured accuracy, so you know what to expect.
03 · BUILD
The model goes into an app, dashboard or backend your team can actually use.
04 · SHIP
We deploy, watch how it gets used, and fix what real users run into.
A core team across computer vision, generative AI, language models and autonomous agents. We start from the real-world problem and engineer backwards to the model.
Sets technical direction across the studio and leads computer vision and product delivery, including HujraCart.
Leads deep learning and generative AI work, from model design to production-ready pipelines.
Builds language-model features into products: retrieval, evaluation and reliable LLM integration.
Designs agents that plan, call tools and complete multi-step work end to end, backed by solid ML.
Tell us what you're working on. The first conversation is free, and we reply to every message.
khansai.contact@gmail.com