Software Design & System Planning
Architecture, build-versus-buy, and the technical decisions that are expensive to reverse once they are shipped.
Fractional Technology Leadership
Senior technology leadership on a defined timeline.
I provide fractional technology leadership to small and mid-sized companies for a defined stretch with objective senior judgment without the permanent headcount. I show you how AI can catapult your business operations. The engagement runs end to end: assess what's working and what isn't, identify ways in which AI can help, and then produce a roadmap for integrating AI in sustainable and meaningful ways. Extensive AI experience and fifteen-plus years leading technical programs across digital health, EHR, clinical trials and federal, including a 115-person business unit and technology portfolios above $10M.
If "How do I use AI" is the question you cannot answer — most companies know AI should be doing something for them, but don't know how or where to start. I find the places it genuinely helps, build those, and leave the rest alone — with human review in the loop and sensitive data handled properly.
In Their Words
Working with Belmont Group has been a huge asset to my consulting business. As a solo consultant working across multiple schools, districts, and organizations, I have a lot of moving pieces to manage. Belmont helped me think through where technology could simplify that work and then actually built solutions around the way my business operates.
The result has been less time spent on repetitive administrative work, better systems for managing and delivering my work, and more time for the things my clients actually hire me to do. What I appreciate most is that they don’t start with the technology. They take the time to understand the problem first and then build something that genuinely makes the business work better.
How I Work
I start by listening. In one week I sit down with you and/or your critical stakeholders to understand the current state, the intended direction, and where friction could be alleviated with AI.
What I produce first is a written account of what I found and where the opportunities are. Some clients want the roadmap and nothing further. Others want me to train their team and work with them — keeping everyone aligned, updating and course correcting the roadmap along the way.
Occasionally a client will want me to just build the solution for them, in which case I can do that. Whether I build it myself or I can make introductions to engineers that can help build the solution. The primary offering is leadership and guidance; building is an optional add-on.
Roadmap Definition
Sequenced over a practical three-to-nine-month horizon, with the trade-offs made explicit.
Business Alignment
So the engineering roadmap and the company's strategy remain the same document.
Team Health & Alignment
Morale is a delivery variable, not a soft one. Keeping a team pointed the same direction — and willing — is most of what makes a roadmap survive contact.
AI Enablement
A substantial part of the practice is AI enablement in internal operations — identifying where it genuinely improves how a company works, then integrating it sustainably, with human review in the loop and sensitive data handled properly. I have built and shipped production tools this way myself, so the guidance comes from having done the work rather than having read about it.
Background
Fifteen-plus years leading technical programs across digital health, EHR, federal, and clinical trials — including functional management of large distributed teams and portfolios above $10M, grounded in deep technical program management experience.
- Ran a $10M+ capital portfolio and a $1.5M operating budget across a business unit at a global clinical research organization, with an organization of 115 full-time and contract staff.
- Built the PMO practice — documentation standards and phase-gate review — across 50+ concurrent projects.
- Held a $9.6M P&L in digital health and grew the underlying contract by $1.6M through a facilitated amendment process.
- Led delivery of a federal grant-disbursement platform replacing a legacy system, distributing $50M annually to roughly 8,000 clinicians under HHS regulation.
- Doubled a chronically over-utilized team, reducing attrition and burnout during a post-acquisition restructure.
- Introduced infrastructure-as-code practice at an EHR provider serving hospitals worldwide, increasing software installation cadence fivefold.
- Recovered an ~$8M/year federal engagement that was at risk of contract loss.
- Migrated 1.2M+ documents in an enterprise document-management modernization.
- Applies LLM-assisted pattern matching over large de-identified clinical datasets to surface trends a single reviewer would miss, with structured human review on every deliverable.
- Has built and shipped production AI tools directly — a custom CRM and internal operations utilities — not only advised on them.
- Executive advisor to client leadership, translating clinical and business needs into technical roadmaps.
- Directs multi-vendor initiatives, aligning roadmaps, dependencies, and SLAs across partners.
- Federal client relationships including the FDA; experience across HIPAA-relevant clinical environments.
- M.S., Management of IT, University of Virginia (McIntire School of Commerce); B.A., Linguistics, UVA.
- Career spanning Sapient Government Services, Starwood Hotels, CliniComp International, ICON plc, Safe Health Systems, and Crowd Tech Group.
- Fluent in English, French, and American Sign Language (ASL).
Engagement Models
Most engagements start the same way — a short, fixed-fee assessment that stands on its own. What follows is your call.
Step One
The Assessment
Brief, focused and structured conversations with you and each of your senior staff, then a written report of what I found — the surprises included — and where the leverage is. I walk you through it in a readout your team is welcome to attend. You leave with a current-state assessment and a prioritized roadmap you can act on immediately. It stands on its own, and nothing about it obligates you to what comes next.
Step Two, if it’s useful
Ongoing Engagement
A set number of days each week alongside your team, working the roadmap the assessment produced. A fraction of the cost of a full-time technology executive, with none of the search and none of the severance — and you can end it at any time. When you are ready to put a permanent leader in the seat, I will help you hire them.
Company Fit. Companies with an established product or operating business and an interest to catch the AI wave but don't know how to.
Not a Fit. Pre-product companies looking for a technical co-founder; organizations that want an outsourced development shop; situations where the decision has already been made and outside validation is the actual request.
Investment.
- The assessment is a fixed fee, most assessments fall between $1,500 – $8,000 depending on organization size and number of stakeholders. Quoted before we start, no surprises.
- Capacity: I hold 2 - 3 concurrent engagements so each gets real attention.
Track Record
Recent engagements. Client names are withheld by agreement; the outcomes are not. Metrics below are client-verified and sanitized. References are available at proposal stage.
Career Record — Leadership
Fifteen-plus years leading the teams that deliver
Before Belmont Group I ran the work from inside. At a global clinical research organization I led a business unit of 115 full-time and contract staff, a $10M+ capital portfolio and a $1.5M operating budget, and built the PMO practice — documentation standards and phase-gate review — across 50+ concurrent projects. I doubled a chronically over-utilized team through a post-acquisition restructure, cutting attrition and burnout, and recovered an ~$8M/year federal engagement that was close to contract loss. An M.S. in Management of IT from UVA's McIntire School put formal structure under what the roles taught.
What it means in practice: the judgment I bring to your team belongs to someone who has held the P&L, carried the headcount, and kept people aligned and willing while the plan changed.
Anonymized Engagement — Healthcare Practice
AI utilities for an independent nurse practitioner practice
A nurse practitioner-led practice running a non-standard care model faced a choice between outsourcing back-office work and adding administrative & support nursing headcount. I mapped where the time actually went, then designed and built AI-assisted utilities for the highest-volume steps — with human review on every clinical-adjacent output, and patient data held inside their own environment rather than a third-party platform.
Result: Approximately 12 hours per week returned to a solo practice’s clinical staff, allowing the practice to defer a planned support-nursing hire and avoid the associated annual employment cost. No PHI left the practice’s control.
Anonymized Engagement — Education
Technical guidance and AI solutions for an education consultancy
A small education consultancy had more internal platform ideas than capacity to build them, and no shared basis for choosing among them. Working with their leadership, I scored each candidate against effort, reversibility, and revenue proximity, then sequenced the two that moved time-to-market and marketing reach — and explicitly deferred the rest.
Result: 3 proposed builds reduced to 2 funded ones; the first shipped in 2 weeks. Deferred items were documented with the conditions that would justify revisiting them.
Built, Not Only Advised
I do not sell software development. But when an assessment finds a gap that is cheaper to close than to specify, procure, and wait for, I can close it — and having shipped production systems myself is why my architecture and build-versus-buy advice is not theoretical. These are examples of that work, shown rather than described.
Built and in daily use
A lightweight CRM
A lightweight CRM - Built and in daily use
The problem. A bootstrapped startup was struggling to keep track of all the leads they were making but could not afford to pay a monthly fee for a CRM that was overloaded with features they didn't need.
The decision. Buy was evaluated first, however the subscription cost was not in the budget. Building was justified only because the required scope was genuinely narrow and the alternative was expensive indefinitely.
What it does. Tracks contacts through a lifecycle; keeps every note, email, and meeting against the record; surfaces a daily working queue; posts activity into Slack. An admin layer handles custom fields, deduplication, audit history, and backups. An AI layer drafts outreach and summarizes the day, with a human approving anything that leaves the building.
Adoption and result. All startup team members involved with outreach and lead generation using it on a steady and regular basis since launch.
Ownership and support. The client owns the source and the infrastructure. Handover included documentation; ongoing maintenance is a defined arrangement.
Built and in daily use
Complex Excel workbooks
Not every problem deserves an application. When the real need is a model a controller can open, audit, and change without a ticket, the right answer is a well-built workbook — versioned, documented, and structured so it survives the person who built it. Choosing this over a platform is often the more senior decision, and it is one I make regularly.
Proof of concept
A custom workflow tool
The premise. Most teams route approvals and multi-step requests through email threads and spreadsheets. Nobody can say who is holding what, deadlines aren't enforced, and there's no honest answer to "where is this stuck?" The off-the-shelf answer is a BPM suite priced per seat and per process, or a heavyweight install. I built this prototype as a proof of concept for myself that a genuinely usable version of that — designed by the operations people who own the process, not by a developer — was feasible at a scope a small team could actually own.
Its features include. A visual editor where a process is drawn as a map: human tasks with their own forms, parallel splits and joins, and decision gateways that route on the request's own data. Publishing freezes a version, so a request already in flight never changes shape underneath it. At run time, work lands in the right person's inbox — assigned to a named individual, or offered to a role and claimed — with per-step deadlines, reminders, escalation to a second person, and an admin force-advance for anything genuinely stuck. File attachments ride along with the task. Every running instance has a live map showing where it is, which branch it took and which it skipped, backed by a full event timeline. An analytics layer renders that same map as a heat map: where work waits, where it runs late, where it escalates, and which gateway branches actually get used in practice. It's multi-tenant throughout — organizations, roles, invitations and permission levels, with each org's data walled off at the query layer rather than by convention.
What it demonstrates. That the interesting engineering in this category isn't the form builder — it's the state model. A process is stored as one immutable JSON definition per published version, and a running request is a set of live branch tokens, so parallel work, skipped paths and force-advanced steps all render correctly on the map instead of collapsing into a single "current step." That's the part that's hard to retrofit, and it's the part I wanted to prove out.
Common Questions
The questions that come up before a first conversation, answered here rather than left to the call.
With the assessment. It is a short, fixed-fee piece of work that stands on its own: conversations with you and each of your senior staff, a written read of what I found, a prioritized roadmap, and a readout your team is welcome to attend. Nothing about it obligates you to what comes next, and a good number of clients take the roadmap and run it themselves.
The assessment is a fixed fee, scoped to the size of the organization and the number of stakeholders, generally somewhere between $1,500 - $8,000. It is quoted before we start and won't change unless mutually agreed upon. Ongoing engagement is priced by a set number of days per week — most clients land at one or two — billed monthly and cancellable month to month. There is no search fee, no equity, and no severance.
One to two weeks from kickoff, typically. What I need from you is a few hours of your own time, an interview with each of the people closest to the work, and access to whatever documentation and systems pertinent to the task at hand.
It fits companies where AI isn't yet being leveraged or business units where it isn't known how to leverage AI adequately. It is a poor fit for a pre-product company looking for a technical co-founder, for anyone wanting an outsourced development shop, and for a decision that has already been made and wants validating.
An executive. The engagement is leadership: architecture and build-versus-buy judgment, roadmap definition, and keeping the roadmap and the business strategy the same document. Occasionally an assessment surfaces a gap that is faster to close directly than to specify, procure and wait for — and in that case I can build the thing. Having shipped production systems myself is why the architecture advice is not theoretical, but building is an option inside the engagement, never the point of it.
Under your NDA or mine, signed before the first substantive conversation. Where an engagement touches protected health information I work inside your environment and your controls, under a Business Associate Agreement where one is required. The default is that sensitive data does not move: tools built for clinical clients are self-hosted in the client's own infrastructure, no PHI is sent to third-party AI services, and a person reviews every AI-assisted output before it is used.
You (the client) do — source, data, and infrastructure, delivered into your repositories and your accounts. There is no license to renew and no dependency on me to keep it running. Tools are handed over documented and deployable, and ongoing maintenance is a stated arrangement rather than an assumption.
Deliberately one person: you get the senior person on every call, and nothing is delegated to someone else. The safeguards are structural. Deliverables land in your systems as they are produced rather than at the end, so at any point you already hold the assessment, the roadmap, the source and the documentation. For work outside my depth — a formal security audit, specialized regulatory or platform work — I say so and bring in a specialist you approve. And ongoing engagements are month to month, so you are never locked into a bet on one person's calendar.
Yes. References are available upon request.
Let's Talk
If your organization is weighing a technology decision that warrants senior judgment, I'd welcome the conversation. And if someone in your network is wrestling with one, I'd appreciate the introduction.
Connect on LinkedInThanks — I'll be in touch shortly.
In the meantime, feel free to grab a slot directly on my calendar.
Book a Time with Mark