lippincott.space
Brandon Lippincott

Brandon Lippincott

Software engineering & systems architecture

Engineering · Architecture · Applied AI

I architect and
build software.

Cloud architecture, cybersecurity, full-stack development, and applied AI.

I’ve spent more than twenty years working across technology and business. I take projects from the initial requirements through architecture, development, and launch.

An architectural model of connected stone platforms, green glass rooms, and garden terraces: a visual metaphor for thoughtful systems.
FIG. 01 / CODE COMING TO LIFE
From technical decisions to working software.
20+ years

Working across technology and business

Architecture & security

Cloud systems, cybersecurity, and technical decisions

Software development

Full-stack applications, applied AI, and product delivery

01 / Selected work

Recent software
projects.

A few case studies on personal passion projects.

  1. AI search · Education BibleVerse Scripture memorization and meaning-based search across 64 editions.
  2. Web platform · Privacy SmarterSources 180+ free tools that run entirely in your browser. Files never leave your device.
  3. Health · Information design Cairn Navigator A plain-language guide for people facing a new cancer diagnosis.
  4. Clinical NLP · Prototype Clinical Note Highlighter Turns medical language-processing output into a visual review tool for coders.
  5. AI · Non-profit Member Knowledge Assistant An AI-powered knowledge platform for a large non-profit organization.

AI search · Education · Live

BibleVerse

A free site for memorizing and reading scripture. It walks you through a five-step memorization path based on learning science (listen, fill in the blanks, partial recall, first letters, then from memory) and brings verses back for review at growing intervals. Around it is a full reader in 64 editions, search by meaning, and a research hub with an interactive "Map of Meaning" of every chapter.

Explore the engineering

Engineering highlights

  • Semantic search without a vector database. The index ships as a compact, quantized static file, and a serverless function searches it in about 2 ms. Typed references like "John 3:16" are answered directly and never call the model.
  • Model choices made by measurement. I built an evaluation harness to compare embedding models and sizes before migrating to Google's EmbeddingGemma. Top-1 recall rose from 67% to 84%, and the index shrank by a third. Production results matched local runs to three decimal places.
  • Responsible AI by design. Searches that sound like someone in distress now show help resources first, and sensitive verses are never shown without their context. The rules are tuned and verified with their own evaluation set.
  • A reproducible "Map of Meaning". All 1,189 chapters are embedded and laid out in 2D and 3D with a fixed seed, so a rebuild comes out byte-for-byte identical. More than 175 automated checks re-verify every claim the page makes.
  • Licensing rigor across 64 editions. One source of truth lists every translation, and tests make sure no edition claims a license nobody has read. Licensed texts are cached with expiry and withdrawn automatically if a refresh fails.

Web platform · Privacy · Live

SmarterSources

More than 180 free tools that replace paid subscriptions for everyday jobs: image and PDF editing, file conversion, finance and real-estate calculators, tools for students and creators, planners, and games. Every tool runs in your browser. Files are never uploaded, and there are no accounts or limits.

Explore the engineering

Engineering highlights

  • Heavy media work, zero backend. PDF merge, split and compress, video-to-GIF, video-to-audio with a hand-written WAV encoder, HEIC and AVIF conversion, and a custom EXIF reader and remover, all done in the browser.
  • A framework-free component system. One shared module adds navigation, a mega-menu and fuzzy site search to 180+ static pages. Toolkit groupings come from a central tool registry instead of being kept by hand.
  • Automated SEO pipeline. Repeatable scripts generate structured data from each page's real FAQ, pre-render internal links for crawlers and AI answer engines, rebuild the sitemap, and check every page before it ships.
  • Automated assets and testing. Every tool's preview image is generated automatically, and automated test suites catch visual and mobile regressions before they reach users.

Health · Information design · MVP

Cairn Navigator

A free, plain-language guide for people newly diagnosed with kidney or pancreatic cancer, and for the family members caring for them. Tell it where you are, and it lays out the path ahead: what usually comes next, who you may meet, words to know, and a printable list of questions for your care team. Every fact cites a reputable source, mostly the National Cancer Institute, and the site never diagnoses or recommends treatment.

Explore the engineering

Engineering highlights

  • Privacy by design: no protected health information (PHI) collected. There is no backend and no accounts. Whatever you enter stays in your browser, with a one-click "clear your data" option.
  • No AI at runtime, on purpose. To avoid hallucinations in a health setting, the path is built from small, numbered rules, each covered by tests. Nothing a user types is sent to a model provider.
  • Medical content managed like code. Each topic is a typed, schema-checked file. CI checks sources, links, reading level (about 8th grade), freshness, and standard medical codes against official code lists.
  • Clinician review with no server. Reviewers approve, edit, or reject topics in the browser. Each decision is fingerprinted against the exact text and sources they saw, so a later edit voids the old review.
  • Ready for AI assistants. A dependency-free, read-only Model Context Protocol (MCP) server lets assistants like Claude look up Cairn's sourced content, with the same safety notes the site shows.

Clinical NLP · Healthcare · Prototype

Clinical Note Highlighter

Clinical language-processing services return dense JSON: entities, character offsets, confidence scores, modifiers and suggested billing codes. This tool turns that output into something a person can review. It shows the original note with every finding highlighted in place, and a review table for accepting or rejecting suggested ICD-10-CM codes. I built it on a commercial healthcare-terminology provider's entity-extraction API.

Explore the engineering

Engineering highlights

  • The note is the interface. Problems, drugs and tests are color-coded in place. Modifiers like body location, severity and dosage are nested or underlined, negated findings are struck through, and clicking any term shows its codes, confidence and related details.
  • A person approves every code. Each ICD-10-CM suggestion is accepted or rejected in a review table, and only the approved codes are exported, as clean JSON.
  • Live confidence threshold. The API is asked for every match once, and a slider filters the results in the browser, so reviewers can adjust sensitivity instantly without sending the note again.
  • A thin, focused service layer. A small Python service handles every exchange with the provider's API, so the interface stays simple and the integration lives in one place.
  • Demo-ready out of the box. A mock mode replays a recorded response, with character offsets recalculated against the sample note, so the tool can be shown without an account and without sending any data.

AI · Non-profit · In production

Member Knowledge Assistant

A members-only platform for a large non-profit organization with many local chapters. Administrators publish the organization's reference material, and members ask questions in plain language. Answers come only from those documents and cite their sources and page numbers. It has served members in production for more than a year.

Explore the engineering

Engineering highlights

  • Grounded retrieval-augmented generation (RAG). Documents are extracted, cleaned, chunked, and embedded locally. Personal information is redacted before anything is indexed, and answers are limited to the retrieved context.
  • Hierarchical, role-based access. Each member belongs to one local unit within a multi-level structure. Six roles control access, and leaders see dashboards and reports for their part of the organization only.
  • Cost-aware AI. Embeddings run on the server's own CPU, so the only paid AI call is generating the answer. Token use and cost are tracked for every answer, and the deployment was tuned to remove cold starts.
  • Tools for the people running it. Administrators can switch AI models at runtime, curate answers, manage content filters, and work through member feedback, with usage dashboards for leaders at every level.

02 / Why work with me

What I bring
to your team.

I bring more than twenty years of experience across engineering, cybersecurity, and business. My background combines team leadership, enterprise architecture, and commercial software delivery.

Engineering leadership

Experience leading and developing teams.

I have led cross-functional engineering teams, hired engineers, and supported their growth through mentoring and performance development. I bring experience setting technical direction and aligning people around shared priorities.

Enterprise architecture

Technical depth across complex systems.

My background encompasses cloud platforms, cybersecurity, enterprise applications, and production AI. I bring the judgment to connect architecture with business requirements, manage technical risk, and guide delivery in regulated environments.

Commercial delivery

A record of delivering for clients.

I have managed the delivery of more than 100 web applications for commercial clients. My experience includes defining scope, pricing engagements, managing delivery, and maintaining long-term client relationships.

Professional credentials

  • CISSP
  • CCZT
  • DASM

B.S. Business Administration · Summa cum laude

People. Systems. Working software.

03 / Technical range

Technical skills
and experience.

Experience drawn from professional engineering, technical leadership, and independent work.

Applied AI

  • Retrieval-augmented generation (RAG)
  • Embeddings & semantic search
  • Evaluation harnesses
  • MCP servers
  • Responsible-AI guardrails

Cloud & DevOps

  • AWS · Cloudflare
  • Serverless & edge
  • CI/CD pipelines
  • Infrastructure as code
  • Cost optimization

Security

  • Application security
  • Cloud security
  • Identity & access management
  • Secure development lifecycle
  • Privacy engineering

Full-stack

  • Python · Django
  • JavaScript · Node.js
  • PostgreSQL · SQLite
  • Progressive web apps
  • Automated test suites

Product

  • Technical SEO
  • Accessibility (WCAG)
  • Information design
  • Product strategy
  • From idea to launch

04 / Better together

Let’s build something
wonderous together.

The best work brings different strengths together. Let’s explore how my engineering and architecture experience can complement your team.