Custom apps.
Two lanes, one team.
Bespoke AI software, sized to the problem — from a focused app for a small business to a multi-service system for a large one. Same senior team, same grounding, same governance.
— Internal beta. Selected work below is anonymized. Ankush to refine.
Right-sized to where you are.
MSME
Production-grade AI apps for small and mid-size businesses — a working app in weeks, on a lean stack you can afford to run. Chat & lead assistants, consumer planning apps, ERP-lite, WhatsApp collections, small-team ops automation.
Explore MSME appsAI Augmented Custom Dev
Complex, multi-service systems delivered by a small senior team with AI in the loop at every stage — agentic backends, predictive/ML engines, trading & market-data APIs, execution control-towers.
Explore custom devAnonymized case studies.
Real engagements, client identities removed. Outcomes are qualitative or client-stated targets where a project is still in build — not audited results.
Conversational finance analytics on Microsoft Fabric
A global elevator & escalator OEM wanted natural-language finance analytics without replacing its governed SAP + Power BI stack. We built a multi-agent layer on Microsoft Fabric: an AI Foundry orchestrator routing to a Fabric Data Agent over the certified semantic model, with a strict "never invent numbers, always trace to source" governance prompt and a large regression-tested query suite. Reached final-demo stage, integrated into Microsoft Teams.
Forecasting, scenario & root-cause analytics
Same OEM, finance domain: manual monthly FP&A and inconsistent forecast variance. We scoped point-solution AI/ML use cases alongside the existing BI — statistical + AI forecasting with dynamic updates, what-if scenario modelling, variance and root-cause analysis separating internal vs external drivers, and an auto-matching cash-application concept. Delivered as bolt-ons, not a BI replacement.
Sales & inventory analytics agent
A MENA-region FMCG leader's sales and inventory data sat on SharePoint with no unified analytics. We built a Fabric Data Agent + Power BI semantic model behind a Copilot Studio sales agent — intent classification, certified-measure DAX queries, TOPN capping for high-cardinality data, role-based access, and an insight layer that describes only what the data shows. Delivered inside Microsoft Teams; pilot-speed 4–6 weeks.
Automated verification of supplier inspection reports
An incoming-quality team checks supplier First-Article Inspection Reports by eye against the engineering drawing — slow at volume and open to a supplier editing a value to pass a failing part. We built a verify-and-audit pipeline: vision-LLM only reads the PDF/drawing; a deterministic rule engine (not an LLM) does the tolerance and GD&T pass/fail and flags tampered rows — so the compliance verdict stays auditable and never hallucinated.
All case studies anonymized — no client names, logos, or confidential figures. Some engagements are in build/POC stage; outcomes shown are qualitative or client-stated targets, not audited results.
Have something to build? Tell us the problem — we'll pick the lane.
Talk to us