LinkedIn · Isos Terra · Recompass · TENS HQ · Upwork · chris@digitaltreehouse.com

I turn messy operational problems into software that ships and stays shipped. I started in Compliance, learned to build by automating my own bottlenecks in Power Platform and Python, and now design and run full products: multi-tenant SaaS, a pricing engine, a mobile game. The common thread is that every one of them is explicit about what it will not compute, because a number you cannot defend is worse than no number.

Now: Portfolio Tooling & AI Enablement Analyst with BRMi. Prior roles and accomplishments are on LinkedIn.

Freelance and consulting: Power Platform and automation work through Upwork, and AI-automation consulting through DigitalTreehouse. For either, email chris@digitaltreehouse.com. I also offer consulting on implementing AI in daily work and life.

Three products, two of them in one arc. Recompass finds public contract data, and turns that into a capture tool for the nonprofits that compete for those contracts. Isos Terra serves the same nonprofits' workforce plans and compliance needs. The Branch takes the career-decision choices to the macro. Public data, then the organization, then the person.

A workspace for the accommodation and career plans a nonprofit agency under an AbilityOne-style contract already keeps: the supports a person uses on the floor, the site's own record of what changed, the monthly check-in, and the growth plan underneath. Three chairs (HQ, Management, Supervisor) see the same records at the altitude their job needs, and every number on the dashboard opens the screen that proves it.

Status: deployed with hosted identity, Postgres persistence and a proven tenant boundary. Two persistence engines with a parity suite showing they agree. Every record on it is fictional by design for now. AI is confined to an explicit request for a wording suggestion that a human must review and confirm. No model scores a person, decides eligibility, or changes a plan.

	Recompass · capture-plan.vercel.app Federal contract capture for AbilityOne nonprofit teams as one workflow: Find → Explore → Price → Team → Propose. The engines behind it are the previous TENS HQ modules I created as portfolio pieces, I ported them into a single multi-tenant Next.js application with typed contracts between stages. Status: pre-pilot currently. The sections Find through Team are built and tested (2,300+ tests); Propose is a stub, not a proposal writer. Subscription billing rails exist in code but no live transaction has occurred. The data-source gate stays closed until the procurement-list feed is proven, so nothing publishes to production data yet. Private repository.
🌿	The Branch Don't compare choices. Compare the lives they create. A career decision platform for young adults choosing a first serious career and working adults in the process of changing one. It projects the whole life a path creates over ten years (earnings, debt, commute, schedule, family load, reversibility) and shows the exact salary where the ordering flips, with the formula version and every assumption visible. It never declares a best career or predicts success. Status: in commercial build-out. Identity, participant-owned Postgres with row-level security, Stripe checkout and fail-closed entitlements are written but unconfigured; there has been no production release. The public demo routes run entirely on synthetic fixtures. Private repository.
	Scandalot Run for office. Survive the scandals. Claw your way from council to president. A political satire roguelite in Flutter for iOS and Android: campaign management, governing choices, debates, donors, and the scandal that ends careers. Original five-motif score, authored debate trees, save migration across office terms. Status: release candidate. Store listings and privacy package are drafted and internally consistent; the remaining gates are Apple infrastructure, device testing and rights evidence, not code. Not yet submitted.

• Complement Engine (working name "Tell"). A personality assessment that scores how you answer, not what you claim. It puts you in situations, reads the tells in your free-text reply (first verb, whether you moved toward a person or the facts, hesitation, rewrites), then names the gap between your self-report and your behavior. Currently a private working build with a blind two-model report evaluation on real respondents. It is not a validated psychometric instrument and is never described as one.

TENS HQ is where the capture work started: six public Streamlit applications that hand each other typed JSON rather than one app with six tabs. Find expiring contracts (GovCon Recompete Radar), investigate the incumbent (ReconRadar), price the work (FMP Calculator), prove compliance (CMMC Vault), staff it (ROCC), and log the pursue/pass call so judgment can be calibrated later (EDGE, deliberately unbuilt until there is a corpus to calibrate on).

The interesting decisions in that stack were all subtractions. I built a pursuit score and deleted it because it laundered screened guesses into something that looked retrieved. I never published a recall figure for the link engine because the miss set is structurally unobservable. The price calculator refuses to price when an indirect rate is blank, because blank and zero are different things and treating them the same understated a real buildup by millions. The compliance tool's sample org scores 89 and is still "not conditionally ready," because 88 is necessary, not sufficient.

Rescue Ops Workbook is the one other people use every day: a five-module Google Apps Script package running intake and volunteer coordination on a dog-rescue nonprofit's live workbook. Two bugs surfaced only after real coordinators used it, a timezone mismatch and a phantom-row append, and no fixture would have caught either. That gap between defensible and used is the most useful thing in my portfolio to think about. Not public; it runs on a real organization's data.

OpsPilot Command Center is one framework that ranks what to automate, proven across two unrelated operations on a single 0–100 scale.

I specify, decompose and verify; a model writes most of the line-level code. What is mine is the specification, the slice boundaries, the gate structure that decides what is allowed to ship, the adversarial review pass, and the calls about what to refuse to compute or delete. Every project above carries a dated ledger separating what was executed and measured from what is merely claimed, and each repo credits the review passes with the finds rather than me.

The same rule applies to AI inside the products. Where a model appears at all it is opt-in, confined to a suggestion a human edits, and never in the path that scores a person, sets a price, or confers a status.

Power Platform (Automate, Apps, BI, Sharepoint) · Python · TypeScript / Next.js · Streamlit · Flutter · Postgres (Prisma, Supabase, RLS) · Clerk · Stripe · Google Apps Script · Playwright / Vitest

Microsoft Certified Azure AI Engineer Associate, plus additional Microsoft and Anthropic credentials. Full list on LinkedIn.