Tanumay GoswamiI move enterprise automation to the cloud
and keep it running.

The problem

An R&D sample that has to cross a border is not one process, it is four: someone requests it, someone approves it, someone prepares the import and export paperwork, and someone tracks where it actually is. Run over email and spreadsheets, every handoff adds a wait, and the requester ends up chasing status rather than working.

What I did

Built one intake on Power Apps so every request enters the same way, orchestrated the approvals and routing in Power Automate, and used UiPath for the repetitive system entry and document handling underneath. The import and export steps were modelled as part of the same workflow rather than as a manual process bolted on at the end.

The problem

Every new hire triggers a series of milestone events that have to land on the right working day. At 650+ events a month this is both a volume problem and a correctness problem: weekends and public holidays shift dates, and a missed milestone is not an abstraction — it is felt by a person in their first week.

What I did

Automated the full scheduling chain across UiPath, Power Platform, SharePoint and Outlook, with weekend and holiday handling built into the date calculation rather than corrected afterwards. Added real-time exception status so anything that failed to schedule was visible immediately instead of being discovered by the new hire.

The problem

Procurement decisions depend on macro and commodity data spread across public and paid providers, each with its own format, refresh cadence and access method. Assembled by hand, the numbers are already stale by the time the deck is built, and nobody can say with confidence which figures were refreshed this cycle and which were carried over from the last one.

What I did

Orchestrated 27+ recurring reports from 10+ public and paid sources into a single Procurement Intelligence Dashboard, automating retrieval and consolidation so the dashboard refreshes on a schedule instead of being rebuilt each cycle by whoever owns the deck.

The problem

A full sector was running 200+ production automations on a legacy UiPath framework, tied to on-premise Orchestrator and an ageing Windows runtime. The framework was going out of support and the runtime had to move, but the automations underneath it were load-bearing — processes business teams depended on daily. A migration that paused operations was not an option, and a big-bang cutover across 200+ solutions was not survivable.

What I did

Inventoried the estate first, then rebuilt each automation against the Windows-compatible framework and repointed it at Cloud Orchestrator. Work was sequenced by business criticality rather than by ease, so the processes with the least tolerance for downtime moved on a schedule their owners agreed to, with the legacy path kept warm until the replacement had proved itself in production.

The problem

Automation code lived inside Orchestrator and reached production by being published from a developer machine. There was no version history, no review step, and no reliable way to tell which build was live. With six teams shipping across regions, two developers could overwrite each other and only find out when a process failed in production. Each deployment took around thirty minutes of manual work.

What I did

Moved the automation codebases out of Orchestrator and into Git, then built CI/CD pipelines in Azure DevOps to package and promote them through environments. Publishing became a pipeline run against a reviewed commit instead of a manual action, which gave every deployment a version, an author and a rollback point.

The problem

Bot licences accumulate faster than they are retired. Every decommissioned process, every pilot that did not proceed, and every over-provisioned environment leaves a licence attached to it — and licences renew whether or not anything is running on them.

What I did

Audited actual utilisation across the estate against what was provisioned, separated genuinely idle capacity from seasonal and low-frequency workloads, and decommissioned what was confirmed unused. The point was to cut cost without cutting headroom the business would need back.

The problem

A regional RPA team was scaling into a Global Capability Centre serving six sectors. Processes built by regional teams had to become processes a central team could support — which is a documentation and handover problem long before it is a technical one. Undocumented automation transferred between teams becomes unsupportable automation.

What I did

Transitioned processes from development into sustain with complete documentation and support readiness for each, so the receiving team inherited a runnable, diagnosable product rather than a codebase. Onboarded six sector teams onto the shared way of working.

The problem

Automation in production fails quietly. A queue backs up, an exception rate creeps, a scheduled job silently stops firing — and the first anyone hears of it is a business user asking where their report is. Reactive support means every incident is discovered late and diagnosed from nothing.

What I did

Ran monitoring on UiPath Insights dashboards tracking logs, queues and exceptions in real time, so degradation was visible before it became an outage, and owned incident resolution and break-fix through to closure. Standardised the aftermath as well as the response: authored four Knowledge Base articles on PepsiCo ServiceNow and created the RCA template the team now uses org-wide.