AI Portfolio

AI Engineering, Documented by Project

This page presents selected AI engineering projects, from production platforms to smaller applied and experimental builds.

The projects below cover generative AI, computer vision, automation, and applied machine learning. They include client platforms and one internal project. Status is noted for each one.

Aiperion — Generative AI Engine for a Canvas Creation Platform

AiperionDelivered

SiegePal built the core AI engine behind Aiperion, which generates dynamic visual canvases from natural language prompts. The system used the OpenAI API with structured output to produce schema-compliant configurations in real time. As the platform matured, SiegePal migrated the architecture to the Model Context Protocol. This included designing a custom MCP server for tighter integration with the canvas rendering pipeline. SiegePal also built the platform's speech-to-text capability using a self-hosted deployment of OpenAI Whisper. This allows users to interact with the platform by voice.

Technologies / Frameworks
OpenAI API, structured output, custom MCP server, OpenAI Whisper (self-hosted), AWS EKS, AWS Lambda
Key capabilities
Generative canvas creation, voice-to-text interaction, real-time structured output

Shic AI — AI Image Pipeline for a Prompt-to-Product Platform

Shic AIDelivered

SiegePal designed and built the AI-powered image features behind Shic AI. The platform lets users generate custom merchandise from text prompts. The feature set included AI image expansion, targeted in-paint editing, image restructuring, and content-aware find-and-replace. It also included background removal for clean product mockups and AI-generated QR codes for branded packaging. The platform integrates with Shopify for its storefront, with a Django backend handling AI processing and business logic.

Technologies / Frameworks
AI image expansion, in-paint editing, background removal, Django, Shopify (Liquid templating), Google Cloud Run
Key capabilities
End-to-end prompt-to-product pipeline, multi-modal image processing, Shopify integration

ACARS Protocol Fuzzer Using Generative Adversarial Networks

Delivered

SiegePal built a GAN to generate synthetic ACARS aircraft communication messages for security fuzzing. The synthetic payloads were used to fuzz the Communications Management Unit, an avionics component. That unit processes incoming VHF, SATCOM, and HF data link messages. The approach surfaced edge-case vulnerabilities that conventional fuzzing methods can miss. It sits at the intersection of adversarial machine learning and aerospace security.

Technologies / Frameworks
Generative Adversarial Network, TensorFlow, Python
Key capabilities
Synthetic protocol message generation, avionics fuzzing, adversarial testing

Music Generation Recurrent Neural Network

Delivered

SiegePal built a recurrent neural network that generates original musical compositions from a user-supplied set of sample tracks. The model learns the patterns and melodic structure of the training data. It then produces new instrumentals that stay stylistically consistent with the source material. The system supports custom dataset ingestion for different musical styles.

Technologies / Frameworks
Recurrent Neural Network, PyTorch, TensorFlow, Python
Key capabilities
Custom dataset training, style-matched generation

WasteScanner AI — Embedded Computer Vision for Waste Sorting

Delivered

SiegePal built a computer vision system that classifies waste items into recycling categories in real time. The system runs on a Raspberry Pi fitted with a camera module. The classification model was trained using AWS SageMaker. Results are shown through an on-device HTML interface for immediate feedback.

Technologies / Frameworks
Computer vision, AWS SageMaker, Raspberry Pi, Python
Key capabilities
Real-time waste classification, embedded AI deployment

AI Resume Analysis Tool

Internal projectDelivered

SiegePal built an internal tool that automates part of its own candidate screening process. It gathers applicant data for a given role and applies AI-based scoring against the role's technical requirements. It supports initial resume review using consistent technical evaluation criteria.

Technologies / Frameworks
OpenAI API, Selenium, Python
Key capabilities
Automated resume ingestion, AI-based technical scoring

In Development

Not a delivered engagement. This is SiegePal's own product work, still in progress.

SiegePal AI-Driven Cybersecurity Platform

In development — not a completed or generally available client product

SiegePal is building its own AI-powered cybersecurity platform. It is designed to orchestrate more than a dozen specialized security agents across AWS, GCP, and Azure environments. The architecture combines LLM integration with MCP servers to support security analysis. Planned coverage spans vulnerability assessment, threat detection, compliance analysis, and incident response. Current work includes the platform's security architecture: role-based access control, API hardening, and tenant data isolation. It also includes integration with more than 30 third-party security tools. The platform runs on a Next.js frontend and a Python/Django backend, deployed on Google Kubernetes Engine.

Technologies / Frameworks
LLM integration, MCP servers, Next.js, Python/Django, Google Kubernetes Engine, RBAC, MITRE ATT&CK mapping
Key capabilities (planned)
Multi-agent security orchestration, multi-cloud integration, automated threat analysis

Have a Specific Requirement?

Explore how SiegePal can support your AI engineering requirements. For cybersecurity work, see the Security Portfolio.

Schedule a Consultation