Desktop AI suite connecting Blender to LLMs. Uses Docker Compose
to orchestrate Vector DB, backend, and frontend.
Manages ephemeral containers for local MCP servers via Docker API .
A simple, straightforward, minimalistic portfolio — no fluff, no gimmicks. Just real work, practical skills, and genuine technical experience.
Desktop AI suite connecting Blender to LLMs. Uses Docker Compose
to orchestrate Vector DB, backend, and frontend.
Manages ephemeral containers for local MCP servers via Docker API .
All personal projects implement strict automated testing, linting, and docs generation via Actions .
Production B2B platform with semantic caching RAG (Kaelux Neural Agent), New Relic monitoring, Sentry
tracking, CI/CD via GitHub Actions
, and CodeRabbit code review.
Training competitive vertical AI model on collected dataset of frontier model system prompts. Infrastructure: GCP / Colab
using NVIDIA
Spot VMs.
LangGraph based agentic orchestration with official Amadeus and Tripadvisor partner APIs for dynamic MCP tool calling.
Largest synthetic n8n workflow dataset (130k+ workflows) for fine-tuning vertical LLMs. Features archetype-based generation and semantic labels. Also on Hugging Face
.
Custom n8n Interface for AI-driven workflow generation. Trained Qwen-3-Coder 14B on custom high-quality JSON dataset from n8n-workflows-36k. Infrastructure: GCP / Colab
using NVIDIA
Spot VMs.
Python scripting dataset for Blender
used in the ModelForge desktop AI suite.
Hands-on implementation of a secure cloud environment with documented compliance steps.
Full-stack web app hosting via Cloudflare tunnels on custom Pi
architecture with Docker
.
Bad USB implementation with Pi Pico & local SLM hosting on Pi 2.
Fully automated content pipelines for YouTube and Twitter/X
using n8n
and Python
.
Simple 2D game built during High School Python I & II courses.
Locally hosted web page on Raspberry Pi from High School Pi course.