Portfolio

AI software we've shipped, running today.

Not concepts. Not mockups. Each of these is a system we designed, built, and run in production, and each one taught us something we bring to every client engagement.

Case studies

CASE 01

Enovari

Flagship · AI memory platform

Our flagship product: a private AI-memory platform that gives any AI assistant a persistent, personal memory, one that survives across sessions, across projects, and across model vendors. It runs locally as a desktop application for Mac and Windows, so the memory belongs to the user, not to a cloud.

  • Persistent memory, 2,678 durable, searchable memories in production (counted 2026-07-10), outliving any single conversation
  • 15-signal retrieval, full-text, TF-IDF, BM25, PageRank, confidence physics, and co-retrieval learning
  • Multi-LLM routing, the same memory serving Claude, GPT, and local models with no single point of failure
  • Specialized personas, a roster of stored minds, each with its own private memory and voice
Visit enovari.ai →
CASE 02

Code Intelligence

Developer tooling

A code-analysis engine that maps every function, class, and dependency across a codebase, six-plus languages in a single pass. It exists because AI that writes code needs to actually know the code: what calls what, what breaks if this changes, what's dead weight.

  • Whole-codebase mapping, every element and relationship indexed and queryable
  • Impact analysis, "if I change this, what breaks?" answered from real call chains
  • AI-ready output, the map feeds our memory platform, so assistants reason over the real structure
CASE 03

The Orrery

Visualization

Navigate your codebase in three dimensions. Built on the Code Intelligence map, the Orrery renders tens of thousands of code elements as an explorable system, dependencies orbit, patterns emerge, anomalies glow. Architecture reviews stop being archaeology.

  • 3D dependency space, ~500 relevant connections surfaced from 47,000+ indexed
  • Animated flow, watch calls pulse through the system to spot hot paths and dead zones
CASE 04

SkwakBox

Content automation

Seven cooperating agents that create, schedule, and optimize content across ten platforms. One team's voice, carried consistently everywhere it publishes, without a person copy-pasting between tabs at midnight.

  • Seven specialized agents, writing, scheduling, repurposing, and performance-tuning as separate jobs
  • Ten platforms, one pipeline from idea to published post, everywhere
CASE 05

Harbor Receptionist

Applied AI · front desk

An AI front desk for real businesses: it answers, holds an actual conversation, captures the details that matter, name, need, vehicle, callback, and books the meeting. Each business is a cartridge of data, not a rewrite, so a new company onboards by configuration.

  • Converses, not scripts, handles the back-and-forth a form never could
  • Business-as-data, services, hours, tone, and compliance disclosures all live in a per-business cartridge
  • Lead-first design, every conversation ends with captured contact details or a booked slot
CASE 06

ELAGRA

Trading systems · research-grade

A trading-systems research laboratory built in C, Python, and JavaScript, engineered for the one domain where a dishonest system costs real money. It records live markets onto a tamper-evident tape, measures whether an edge actually exists before any strategy may act on it, and connects to a live broker behind hard safety interlocks.

  • Tamper-evident market tape, every tick recorded with integrity checks; the data cannot lie about what happened
  • Edge-measurement gates, no strategy trades until its edge is measured on the tape, never assumed
  • No-lookahead discipline, structural gates make it impossible for a backtest to peek at the future
  • Live broker integration, real order paths behind an owner interlock: live placement requires explicit human arming

Research-grade and internal, listed as engineering evidence, not as an investment product.

Your project

The next case study could be the system we build for you.

Tell us what's eating your team's time. We'll show you what we'd build, starting with a free map of where AI actually fits your business.

Map your workflow →