Custom algorithms & modeling
From messy reality to solutions you can trust.
Models and software for problems too difficult to solve on your own. Then algorithms and applications that help you act with confidence.
Helping you better understand your world.
Work
Here is what that looks like in practice: hard problems, careful models, and software people could put to work.
Letting schedule optimization
For a state transportation agency we assigned a year of construction projects to monthly bid lettings while holding district workload, project mix, and seasonal limits inside planning targets.
- Genetic search over monthly assignments
- Uncertainty analysis for bids and slippage
- Reports and a planning application their staff could run again
Sensitivity analysis
When a model’s answer has to be trusted, you need to know which inputs drive it and how far the answer moves when data shift.
- Trace results back to the inputs that matter
- Test small and large changes
- Show where the result holds and where it moves
Optimization on GPUs
Some problems gain enough structure that GPUs and paired methods change what can be solved in practice.
- Problems selected for the structure the hardware can exploit
- Numerical search paired with AI methods where they help
- Implementations in use and still being extended
Approach
From a hard, messy problem to solutions you can trust: we listen, we model with care, we show the limits, and we ship software people can use with confidence.
Applications
Software we ship for specific domains. Separate product brands; same craft underneath.
AgenSeer
Helping insurance companies better manage their agents’ book of business.
- Agent performance analytics
- Book of business management
- Commission tracking
Workcastr
Capture work on a timeline, build period reports, and track goals tied to projects.
- Timeline of plans and logs
- Period reports you can share
- Goals tied to projects
Stip Letter
Planning application from the letting work — assign lettings and reason about risk.
- Monthly letting assignments
- Uncertainty on bids and slippage
- Practitioner tool staff can run again
About
Some problems are too difficult to solve on your own — too tangled for a spreadsheet and too specific for off-the-shelf software. That is where we work.
We listen until the problem is clear, translate it into a model as accurate as it needs to be, and explain what the model can and cannot do before anyone relies on it. Then we build the analysis and software that put it to work, and we hand back results that tell you what to act on next.
Contact
Happy to field questions about modeling work or the applications.