A practical mind for messy problems
I help businesses fix inefficient processes, connect disconnected systems, and build the software or automation they’re missing.
Sometimes that means AI. Sometimes it means making better use of what you already have. It starts with understanding the problem.
25+ years connecting the dots.
Software / Data / Energy & operations / ConsultingWhat I do
The copy-and-paste routine. The systems that don’t talk. The tool that almost does the job. That’s where I can help.
Moving files. Re-entering data. Running the same reports. If someone repeats the same steps every day, there’s probably a better way.
I look at the whole process, work out what needs human judgment, and automate the rest.
One application has information another one needs. Your team ends up carrying it between them.
I connect applications, databases, files, and workflows so information gets where it needs to go.
Sometimes the software you have almost solves the problem. Almost.
A small internal tool, a dashboard, an integration, or a full application—I build what fits the work you actually do.
Research, document processing, finding answers in your own information: there are useful places to start.
I help you work out where AI adds value, and where conventional software will give you a more reliable answer.
Sound familiar?
They cross departments. They involve several systems. Nobody quite owns the whole thing.
A spreadsheet holding everything together.
A manual process someone runs every morning.
An expensive application everyone works around.
“That’s how we’ve always done it.”
That’s a starting point. I trace the workflow, find the connections, and work out what would actually make things better.
Start with the people doing the work.
Follow the information across systems.
Choose the tools that fit.
Selected work
A few examples from my work. Different problems, different tools. The same aim: make things easier to run, understand, and maintain.
Featured project / Practical AI
A locally running language model connected to the dashboard’s SQLite database, operational logs, and error logs.
The problem. Answering questions about scheduled tasks, scripts, and failures meant looking through database records, execution logs, error logs, and documentation separately.
What I built. An AI assistant integrated into M.A.D.’s Ask MaDi tab. It uses a local language model to answer plain-English questions about the SQLite database and operational or error logs, with scoped tools and curated documentation providing the information behind the answers.
The result. A conversational way to explore operational data and understand logged errors within the dashboard, without first writing a query or hunting through separate logs.
A local language model, SQLite database access through audited read-only SQL tools, and access to operational and error logs. Local model routing, keyword fallback, and curated knowledge retrieval connect questions to relevant data and documentation.
The assistant also supports questions about live scheduled-task status. Administrative controls remain separate from its information-retrieval tools.
Operational software / Automation
The automation was already doing important work. Understanding it, supporting it, and finding what needed attention was the harder part.
The problem. Scheduled jobs, scripts, and logs were spread across different places. Tracing what ran, which scripts it used, and where to look when something failed took too much digging.
What I built. A desktop application that brings scheduled tasks, scripts, dependencies, and logs into one searchable workspace. It also provides file monitoring and tools for adding consistent error handling.
The result. One place to follow the automation from task to script to log, making day-to-day support and troubleshooting easier.
Search tasks and scripts together, see their links, and organize the view around the work.
Inspect script content, logs, and file-processing history without hunting through folders.
Preview logging and error-handling changes before applying them, with separate admin and read-only roles.
The application reads live Windows scheduled tasks and discovers their linked scripts. Re-importing refreshes the catalog while preserving user-edited descriptions and manual links.
FileWatcher starts processing when files arrive, records run history, and provides watchdog recovery and alerts for files that remain unprocessed. Configurable tools prompt for parameters before execution and remember the last-used values.
Each user can save their own column layout and grouping. Administrators can manage tasks and access, and review a diff before applying script changes.
Energy systems / Offshore well data
Independent consulting work involving Chevron, Well Dynamics, Halliburton, and Baker Hughes.
The problem. Offshore well data needed to move through retrieval, calculation, storage, reporting, and downstream systems as one connected workflow.
What I built. Intelligent Well Completion software for Chevron’s Agbami and Tahiti fields, with extensive operator interfaces and automated, redundant SQL Server backends. The systems handled downhole data from collection through reporting and downstream delivery.
The result. An integrated workflow for collecting, calculating, storing, and delivering well data, supported by redundant database infrastructure.
Built with .NET and SQL Server. The related SoftFlow application estimated flow rates and integrated its calculations and reporting into the IWC systems.
Business software / Systems integration
Business software spanning desktop applications, web interfaces, and the connections between them.
The problem. Business administration, inspection applications, and accounting systems needed to exchange information across desktop and web software.
What I built. Multi-user business software, administrative web applications, and integrations for ELD Engineering. The work included a mobile-compatible web rebuild, REST APIs, and synchronization services connecting QuickBooks Online, SnapInspect, and other data stores.
The result. Connected applications and data exchange supporting the business workflow, with implementation, documentation, user training, and ongoing support.
The earlier desktop system was built from the ground up, including requirements, database design, automated migration from Excel, role-based access, and operator event logging. Later work extended the software through web applications and integrations.
FlowCal — Dynamic List Query Engine
The problem. User-defined FlowCal list queries depended on logic normally assembled by the application’s client interface, making them harder to use in automated downstream workflows.
What I built. A query engine that reconstructs the SQL behind those lists, including field handling, joins, logical grouping, operators, and effective-date rules. It stores compiled queries in a shared cache for downstream use.
The result. Reusable query definitions that make FlowCal list logic available beyond the application interface.
PowerShell, Oracle SQL, canonical query generation, hashing, and merge updates.
The person behind SmudgeBrain
Mark Davis / Founder, SmudgeBrain
I’ve spent more than 25 years solving problems with software, data, automation, and systems integration—much of it in energy and operational environments.
I’ve been a developer, architect, consultant, manager, and the person who follows a problem across several systems to find out what’s really happening.
I’m a lifelong, self-directed learner. The technologies keep changing. The part I enjoy stays the same: understanding something complex, finding the connections, and making it work better.
My dad once joked that he hadn’t left his “mark” on the world—just a smudge. That line stayed with me. Now it has a home here.
C#/.NET, PowerShell, Python, SQL Server, Oracle, APIs, desktop and web applications, operational automation, document processing, and local AI systems.
The work spans development, architecture, integration, and support—following the problem wherever it leads.
Let’s figure it out
You don’t need to know whether the answer is AI, an integration, an application, or a 50-line script. That’s part of the job.
Contact details coming soon.
