Duelim combines process improvement, modern data engineering, and AI to replace high-friction manual workflows with governed software that actually lasts.
Handing critical operations to fully autonomous agents is too much of a liability. High-stakes workflows need boundaries, human approvals, and sane failure modes. But operational AI only works as a layer on top of a well-designed process with reliable data underneath it. Most engagements skip the process and data work to rush the automation, and that's why they ultimately break.
I'm Chris Davis. I've spent 15+ years across process improvement, data engineering, and product development. I built Duelim to bring those three disciplines into one delivery model.
Started in manufacturing and supply chain environments at Johnson & Johnson and BAE Systems, then moved into corporate operations roles at National Oilwell Varco and KCI. Led global process improvement projects and rolled out operational KPIs across hundreds of service centers. Over $10M in documented process savings.
Scaled strategic data practices in senior roles at Bloomscape, PeerStreet, Cameo, and HubSpot. Fluent in the modern stack: dbt, Snowflake, DuckDB, Airflow, Hightouch.
Founded Bootstrapital, where I've advised early-stage founders and taken multiple products from zero to one. That 0-to-1 experience led directly to building Rubot, a framework for governed, semi-autonomous workflows, and ultimately to Duelim, where all three disciplines finally operate as one practice.
Every engagement follows the same sequence because the sequence matters. Skipping steps is how AI projects produce demos instead of durable software.
I start with value stream mapping to find where the waste actually lives before writing any code.
I redesign the process itself first, so automation is applied to a workflow that's already sound.
I build auditable data models your operators can trust, using dbt, Snowflake, and the modern stack.
I ship semi-autonomous software on Rubot: workflows that pause for human approval when it matters, with a full audit trail behind every decision.
I scope tightly to one workflow at a time, with explicit success metrics defined before I start.
Every engagement is delivered on Rubot, a framework I built for governed, semi-autonomous internal tools. It is the infrastructure layer that makes those workflows durable in practice.
Rubot gives every workflow durable state, human-in-the-loop approval gates, and a full event trail. When an operator asks why a request was routed a certain way, the answer is in the run history. If a workflow pauses for two days waiting on a decision, it resumes exactly where it left off.
Workflows survive restarts, long waits, and external dependencies. No lost context, no silent failures.
High-stakes decisions route to a named approver. The workflow waits, then resumes on their decision.
Every tool call, model output, and approval is logged to a structured event record. Fully replayable.
A mountable admin surface for reviewing runs, approvals, and workflow state without touching the database.
You work directly with me, not handed off to a junior team. Every engagement follows the same three-phase sequence. Scope is tight by design.
I start with value stream mapping, not technology. I interview the operators, trace the data flows, find the decision points and exception paths, and identify where the waste actually lives. The output is a concrete pilot design with defined success metrics and an honest assessment of data readiness. If automation isn't the right answer for a given workflow, I'll tell you that too.
I redesign the workflow first, build reliable data models your operators can trust, then ship production software on Rubot with approval gates, operator queues, and a full audit trail. One workflow, one operator surface, in production within 4 to 8 weeks.
Once the workflow is live, I track the metrics set in the audit, refine model quality, and expand to adjacent workflows only when the economics clearly support it. No scope creep, no roadmap theater.
Tell me about one operational queue that's costing you time, headcount, or accuracy. I'll give you an honest read on whether process work and AI can help, and what that would look like.
hello@duelim.com · Response within 1 business day