Proof, not promises

AI Automation Portfolio

Explore the AI automation projects we've designed and delivered across content, recruiting, real estate, infrastructure, confidential documents, and multi-agent products.

01Understand
02Design
03Build
04Refine

How to read the labels

Each project carries a status label. Client · Delivered means built for a client and handed over. Internal · In use means built for our own operations and running today. Project · Built means a working system we built that is not a client engagement. Release pending means built, with public release still to come. Some projects were delivered under SiegePal, our cybersecurity and AI security brand. EvoPal and SiegePal are operated by the same company and share one team.

Opportunity screening that keywords couldn't handle

INTERNAL · IN USE
AI-assisted opportunity screening workflow

Our automation screens listings written in endlessly varied terms, recommends promising opportunities, and drafts application material. A person decides what to pursue and submits it.

Explore the project

A content pipeline with no manual steps

INTERNAL · IN USE
Automated document and content production workflow

An AI-powered workflow follows detailed editorial rules to write, illustrate, and schedule content across channels.

Explore the project

The AI engine behind Aiperion

CLIENT · DELIVERED
Structured canvas modules generated by an AI engine

We built an engine that turns prompts into structured canvas configurations in real time, then added a custom MCP server and voice input.

Explore the project

Social content shaped by news, reviewed in Slack

INTERNAL · IN USE
Automated social content cards connected for review

A multi-channel workflow finds relevant news, prepares distinct LinkedIn, Instagram, and Facebook content, and routes the source and drafts to Slack for approval.

Explore the project

Candidate review that turns résumés into a shortlist

PROJECT · BUILT
Candidate profiles ranked for recruiter review

Indeed submissions, résumés, skills, and role requirements are evaluated together to give recruiters a ranked, evidence-linked shortlist.

Explore the project

An MLS deal finder that watches while you work

PROJECT · BUILT
Real estate listings screened on a city map

New MLS listings are screened against a user's investment criteria and promising properties trigger an alert with the source information attached.

Explore the project

GPU infrastructure integrated for TenDev

CLIENT · DELIVERED
GPU infrastructure connected to production AI endpoints

We implemented RunPod integration and deployment for TenDev, connecting GPU-backed AI workloads to a production-ready application flow.

Explore the project

Confidential document processing for Tevot AI

CLIENT · DELIVERED
Confidential documents processed through a secure desktop application

We built a desktop document-processing application in Flutter with a privacy-first architecture spanning AWS Bedrock and Azure AI Foundry.

Explore the project

Multi-agent debate with an impartial AI referee

RELEASE PENDING
AI agents submitting arguments to an impartial referee

FightKlub brings several AI agents into a structured debate, then uses a separate referee to evaluate and rank the responses. Release pending.

Explore the project

AI product creation for Shic

CLIENT · DELIVERED
AI imagery transformed into finished custom products

We built the image intelligence behind a prompt-to-product experience, including expansion, in-painting, restructuring, find and replace, background removal, and AI QR generation.

Explore the project

Other engineering work

The team's wider AI work includes computer vision and custom model training. One example is an image classifier trained with AWS SageMaker. It sorts waste items into recycling categories and runs on a Raspberry Pi with a camera. It isn't business process automation. It does show that we work below the API layer when a project needs it.

Have a process that belongs on this page?

Tell us about the workflow. If automation fits, it could be the next project we write up here.

Walk us through a process