At BTI360, the argument over who owned the office ping-pong table acquired a software solution. An engineer named Tyler built an app using chess’s Elo rating system. Matches became data. Players acquired rankings. Winners even got personalized victory songs. It is a small, pleasingly excessive answer to a familiar workplace problem: everyone has a theory about who is best, and nobody agrees on the evidence.
- Custom software, search and AI for federal missions.
- Horizon tools turn expert rules into reviewable workflows.
- Secure deployment and human judgment shape the design.
That little leaderboard offers a useful entrance into a company whose main work happens in considerably less playful settings. BTI360 builds software for federal government customers, with a particular emphasis on national security. Its territory includes finding meaning in sprawling datasets, moving applications into secure infrastructure, and helping analysts spend more time interpreting information. The recurring question is simple: how do you make an answer useful enough to act on?
A bad project, then a phone call
Co-founder MJ Wivell traces the business to an experience on a $50 million software project. In his telling, the contractor lacked direction and capable technical leadership while continuing to collect payment. On his honeymoon, a warning about entrepreneurship in Rich Dad Poor Dad struck him as an invitation. He called childhood friend Jeremy Nimtz. They founded BTI360 in 2004.
The interesting part of that origin is the diagnosis. Wivell’s complaint concerned how a project was led and staffed. The response was to build a company around developing engineers. Today, Wivell is CEO; Nimtz is controller. Their company’s stated priorities still include technical growth and helping engineers become effective teammates and leaders.
The analyst keeps the red pen
Consider a document reviewer. A keyword can locate a word; it cannot, by itself, decide what a passage means under a particular policy. Horizon Highlight, one of BTI360’s AI products, applies configurable analytic rules to documents. It can propose highlighting, redaction or flagging, while showing the supporting text and reasoning. The analyst can accept, reject or change the recommendation.
That matters because expertise often lives in exceptions: the passage that looks harmless until someone understands its context. Highlight lets teams express their judgment in reusable rulesets, revise those rules and apply them across a collection. Its advertised uses include classification review, document redaction and high-volume triage. The product is designed for secure, air-gapped settings, where the system operates without depending on external connections.
Horizon AgentFlow tackles the sequence around those decisions. Its visual canvas lets users assemble multi-step analytic workflows without writing every component from scratch. Evaluation, provenance and human review are part of the design. Highlight applies document rules; AgentFlow coordinates broader work. Together, they show where BTI360 is placing its product bet: repeatable analysis with inspectable steps.
Conceptual illustration of Horizon’s approach.
The model has to fit the room
A secure environment changes the economics of AI. Computing capacity is finite. Data may need to stay inside a disconnected network. A workflow that performs beautifully through an outside API can become awkward when that connection disappears. BTI360 explicitly designs around these restrictions.
Its approach starts by proving a useful capability with a large language model, then considering a smaller, mission-specific model for deployment at scale. The company reports an 11-fold inference-speed improvement and 91 percent lower cost in its own comparison, with comparable quality. Those are company-reported results, dependent on the workload and deployment. The transferable idea is the order of operations: establish the task’s value, measure quality, then optimize the model.
The constraint also defines the fit. Narrower models make sense when a task is sufficiently understood to specialize and evaluate. A shifting, open-ended assignment may still call for a broader model. BTI360’s own offering spans large models, small models and classical machine learning. Choosing among them requires an engineering decision about the actual job.
The unglamorous part is the product, too
BTI360 also sells the work around the tools: advisory services, custom development, data engineering, cloud engineering and software delivery. Its background includes distributed search and open-source technologies. It describes itself as a certified AWS partner. Buyers are organizations commissioning capabilities and integrating them with existing systems; the engagement extends well beyond opening an account and typing a prompt.
That places the company between specialist AI vendors and the wider federal software-services market. Its pitch combines domain knowledge, product components and the engineers who can put them into operation. For a buyer comparing alternatives, the useful test is practical: can the supplier integrate the data, preserve permissions, explain the output and maintain the application after deployment?
“You build it, you run it”
BTI360’s stated engineering principle
Its September 2026 engineering series makes that last obligation concrete. Developers own automated tests and share responsibility for code. Teams record architectural decisions, including rejected options and trade-offs. They also deploy, monitor and support their applications. Those habits are available to any software organization willing to budget time for them. They make the people writing a feature accountable for what happens when somebody uses it.
A classroom with room for children
The investment in people has physical form. In 2021, BTI360 opened a 30,000-square-foot Herndon office with space for internal classes and technical events. It also included a sound-insulated children’s room beside hot desks for parents. The detail is revealing: a workplace designed around learning still has to accommodate the rest of a person’s life.
Training is similarly specific. Its published offerings include a two-day Lean Agile course, six sessions of hands-on clean-code instruction and four sessions on collaboration. Communication gets teaching time alongside programming. A code review, after all, only improves software if the people involved can make themselves understood.
The business is extending that logic into new markets. BTI360 announced its first advisory board in September 2025, including former DIA acting director David Shedd and former CIA digital-innovation leader Jennifer Ewbank. That December, it announced AxioSight, a commercial spinout initially targeting legal work. The common thread is recognizable: professionals facing too much information, consequential decisions and a need to inspect the reasoning. Even the ping-pong players wanted that much.