Let AI do the busywork.
Keep people on the decisions.
I build AI workflows and agents that read, sort and act on your data. Rules where rules work, AI only where judgment is needed, and a human approval step for anything risky.
Teams usually come to me with one of these.
If your people spend the day copying, checking and sorting, that is work AI can take on.
Four steps from stuck to shipped
Map the work
Which steps are rules, which need AI, and where a person signs off.
Build the pipeline
Collection, extraction and checks, with confidence thresholds.
Add the actions
Agents that act through your tools (CRM, email, databases) behind approval rules.
Run & tune
Monitoring and a review queue as volume grows.
Know what to automate before you pay to build it.
A step-by-step plan for your workflow, with the rules, the AI and the approvals marked, and a fixed quote.
He has driven the project with autonomy and has a solution focussed mindset towards creating unique, cost-saving and data-smart solutions.
Pesql · AI · Data · 5.0 Upwork reviewQuestions before we start
Yes, where cost or privacy calls for it. I have run local models in client work, Qwen 2.5 7B through Ollama and Phi-3 through llama.cpp, so there is no per-request bill from a hosted AI provider.
Rules first, AI second: rules handle the routine cases and a model only the judgment calls. A rules-first, AI-second design cut one client’s running cost by roughly 75%.
Checked output, confidence thresholds and approval rules. Anything uncertain goes to a review queue for a person, and risky actions wait for sign-off.
Yes. All code and IP transfer to you on final payment.