Tampa, FL · Enterprise RPA & AI automation
Nine years of Blue Prism, UiPath, and AI-driven workflows for teams who need a manual process to just run — every day, without anyone babysitting it.
All work contracted through Upwork — escrow, verified payments, and 41 client reviews on the record.
| Process | Runs | Exc. | Avg | Last run |
|---|---|---|---|---|
| invoice-intake | 1,284 | 3 | 42s | 03:00 ✓ |
| rate-change-sync | 221 | 0 | 2m 11s | 02:15 ✓ |
| oracle-db-cleanup | 365 | 1 | 6m 04s | 01:00 ✓ |
| email-routing | 8,940 | 12 | 1.2s | 03:12 ✓ |
| statement-recon | 52 | 0 | 4m 38s | queued |
Andrew Randolph
Automation Engineer · RPA Developer
AI Workflow Automation
Tampa, FL
The journey
I started in RPA in 2016 at Liberty Mutual, building front-office bots against claims workflows. Three bots eliminated the secondary tasks of seven full-time roles. That first project taught me the lesson the rest of the career has been built on: the hard part of automation is never the happy path — it's what the bot does at 3am when the screen it expected isn't there.
At Accenture from 2019, the work got broader. Manufacturing, utilities, financial services, banking — multi-client enterprise RPA across industries that share almost nothing except rules-heavy processes and a low tolerance for failure. A specialty metals manufacturer, a credit-union service organization, an investor-owned utility, an investment bank. I earned the Blue Prism ROM Architect certification in 2019 and Certified Professional Developer in 2021, and spent as much time on solution design documents and requirements discovery as on building.
Since 2023 at SS&C, the work has gone deeper rather than wider: owning delivery end to end, and extending Blue Prism with custom C# and VB.NET code stages where the native components run out of road. Alongside it I've been taking independent work on Upwork — smaller businesses, lighter tooling, Python and n8n and Zapier instead of a full enterprise platform.
In 2025 I started building my own product: a multi-tenant SaaS platform for missed-call lead recovery, serving home-services and restoration businesses. Python, Claude API, Twilio, Stripe, PostgreSQL — six production phases behind 500+ automated test assertions. It's the clearest statement of what I actually believe: the discipline that makes enterprise RPA survive production — structured exception handling, work queues, retry logic, real monitoring — is exactly what LLM-driven workflows are missing. AI changes what a step can do. It doesn't change what it takes to run that step ten thousand times without someone watching.
Selected impact
Every automation gets justified the same way: what it costs to build against what it gives back each year. These are figures from delivered work.
Seven production automations for a specialty metals manufacturer, including Oracle DB cleanup and email-routing workflows.
Three front-office bots supporting claims workflows at a national insurer, eliminating secondary claims tasks.
Reconciliation, rate-change, and dialer-management automations for a large credit-union service organization.
Oracle database cleanup identifying and inactivating unused formulas, routers, and item numbers.
Sales-order rejection email processing, routing, and archiving — a queue that used to be worked by hand.
Maintained and enhanced for an investor-owned utility: four new builds and ten enhancements to a live estate.
Services
I start with the process, not the tool. Some of what lands in my inbox is better solved with an API call than a full RPA build — and I'll say so before you pay for it.
Blue Prism and UiPath built as systems, not scripts: work queues, structured exception handling, credential management, reusable object design, and architecture that still works after the pilot becomes 40 processes.
Claude API integration in production, LLM-driven workflow design, and honest evaluation of where an agentic pattern beats deterministic RPA — and where it plainly doesn't.
REST API integration, Oracle and PostgreSQL database automation, web and data scraping, email and document processing. Connecting the systems that never got an API.
Inheriting fragile bots that fail nightly: root-cause analysis, long-term fixes, and the runbooks and support procedures that let your own team keep them running.
Selected work
Clients are described by sector rather than named. Happy to talk specifics on a call.
Specialty metals manufacturer
1,300+hours saved per year
Designed and deployed seven automations into daily production, including an Oracle database cleanup routine and email-routing workflows that had been handled by hand. The estate ran unattended on a schedule, with structured exception handling that escalated to a person only when something genuinely needed judgment.
Investor-owned utility
14processes maintained
Refactored password-rotation processing onto work queues so multiple bots could run it in parallel, then maintained and enhanced fourteen production processes for the utility's security and logical access team — four new builds and ten enhancements to an estate already in service.
Credit-union service organization
192–960hours saved per year
Led reconciliation, rate-change, and dialer-management automations for a large service organization supporting credit unions — rules-heavy, deadline-bound work where a missed step is a compliance problem rather than an inconvenience.
Investment bank
Year 1return on investment
Co-led a critical investment-bank automation achieving ROI within the first year, authored the solution design documents, and drove the Phase 2 enhancements the bank's own team worked from afterwards. Documentation that outlives the contract is part of the deliverable, not an extra.
Own production SaaS
500+automated test assertions
A missed-call lead recovery platform for home-services and restoration businesses, delivered across six production phases. Python, Claude API, Twilio, Stripe, PostgreSQL — with multi-tenant data isolation, payment processing, telephony integration, and production monitoring. The Claude API handles conversation; the exception handling, retry logic, and monitoring around it came straight from enterprise RPA.
Back-office finance
100%processing accuracy
Designed, tested, and deployed three back-office bots covering invoice cashing, cash balancing, recurring reporting, and email notifications — built from reusable components, and run with no loss of processing accuracy against the manual baseline.
How I work
Requirements discovery with the people who run it today — SME workshops, process mapping, volumes and time spent, and where it currently breaks. Automating a bad process just makes the bad outcome arrive faster.
A ranked recommendation with effort against annual hours saved. If a step is better handled by an API call, a short script, or a change to the process itself, that is what I'll propose.
Structured exception handling, logging, runbooks, support procedures, and knowledge transfer. You should not need me on retainer to keep your own automation running.
Background
Nine-plus years delivering production automation across financial services, insurance, utilities, manufacturing, and technology — before any of it became a freelance practice.
Certifications
Stack
Based in Tampa, FL. US hours, US contracts, no timezone gymnastics.
Next step
I'll tell you straight whether it's worth automating, what it would take, and what it would save. I'm not the cheapest developer on Upwork — I'm the one clients bring in when the last build didn't stick.
Contracts, escrow, and reviews all handled through Upwork.