Case Studies
Every engagement below started with a business constraint — an ERP that couldn't scale, a cloud bill nobody could explain, an AI pilot that never shipped. Here's what happened when we got involved.
A precision engineering MSME running production on spreadsheets and a legacy Tally setup. No real-time inventory, no BOM-driven planning, month-end close taking 14 days. We deployed a composable cloud ERP covering finance, production planning, and inventory — live in 14 weeks. Inventory reduced ₹1.1Cr in the first year through demand-driven replenishment. Month-end now closes in 4 days.
A consumer goods distributor selling across 3 marketplaces, their own website, and 40 retail partners — with inventory managed in a manually updated spreadsheet. Overselling 80+ units per month. We built a unified inventory layer connected to all channels in real-time, with ATP-driven allocation rules and automated replenishment triggers. Stockouts on core SKUs dropped 73%. ₹85L in redundant safety stock was released as working capital.
A 60-person IT services firm losing 35% of billable capacity to miscoded timesheets and unbilled work. Finance closing books in 12 days. AR collections manual and inconsistent. We implemented a project-based ERP with integrated time capture, automated billing triggers, and a dunning workflow. ₹1.2Cr in previously un-invoiced work recovered in year one. DSO dropped 18 days. Book close now 5 days.
A payments fintech with a ₹28L/month AWS bill and no cost governance. Idle instances, dev environments running 24/7, and every workload on on-demand pricing. We ran a four-week audit, rightsized 60% of instances, implemented automated dev environment scheduling, moved stable workloads to reserved pricing, and deployed tagging and budget governance. Monthly bill dropped to ₹18.4L — ₹9.6L saved per month — with identical performance and improved reliability.
A SaaS platform on a monolithic cloud-hosted architecture — traffic had grown 8× in 3 years but infrastructure cost had grown 11×. Manual server provisioning before traffic spikes, 12% utilisation post-spike. We extracted core services into containers, deployed on Kubernetes with horizontal autoscaling, implemented a CI/CD pipeline, and moved to infrastructure-as-code. Infrastructure cost per user dropped 61%. Deployments went from monthly to daily. The next Black Friday required zero manual intervention.
A logistics company's core booking system was a 2007-era application running on a server nobody was allowed to move. One engineer understood the codebase. Every feature request took 3× longer than planned. We applied the strangler fig pattern — building new services alongside the legacy system, routing traffic module by module until the legacy was handling nothing critical. Completed in 9 months. The legacy system was decommissioned with zero production incidents. The new architecture supports feature deployment in days, not months.
460 documents per week — supplier invoices, customs declarations, delivery notes — processed entirely by hand across a 4-person team. 29 errors corrected per week. 3-day average processing lag. We deployed an IDP system combining OCR, vision models, and validation rules against the ERP. 91% of documents now post to the ERP without human touch. Processing time dropped from 3 days to under 2 hours. The team of 4 now handles exceptions only — and manages 30% more document volume with the same headcount.
A demand forecasting model that had been "nearly ready" for 14 months — accurate in a notebook, never deployed. The data science team had the model. Nobody had built the production infrastructure around it. We took over the engineering: rebuilt the data pipeline with validation and drift detection, containerised the model, wired it into their ERP via API, and deployed with monitoring. Live in 11 weeks. Six months later — 22% reduction in excess inventory, near-zero stockouts on the 40 highest-value SKUs. The model retrains automatically on new transaction data weekly.
A procurement team of 4 managing 340 suppliers and 1,200 POs per month — at capacity, unable to absorb growth without additional headcount. We deployed an agentic AI procurement assistant: automated PO generation from approved requisitions, three-way invoice matching, standard supplier query handling, and contract expiry alerting with draft renewals. One year later, the same 4-person team manages 520 suppliers and processes 1,900 POs per month. Headcount: unchanged. The team now focuses on supplier negotiation and strategic sourcing.
Your turn
Every case above started with one honest conversation about what wasn't working. Book 30 minutes with our engineering team — no pitch, just a technical discussion about your specific situation.