Selected screenshots and explanatory illustrations of my personal projects and professional team contributions. My scope is stated for every case. This collection does not include full repositories, model weights or private client material.
Gymnasia
Personal fitness application with a conversational coach and a food-vision agent. I built local tools, streaming tool-use, provider adapters and barcode lookup. The project demonstrates agent design and integration; it is still in development.
https://gymnasia.maximofn.com/
https://github.com/maximofn/gymnasia
LangGraph Deep Researcher
Personal research application built with LangGraph. A supervisor distributes research across parallel agents, compresses findings and produces a sourced Markdown report. I built the Python/FastAPI backend, React/TypeScript interface and live WebSocket progress, with configurable model providers and user-supplied keys.
https://deepresearcher.maximofn.com/
https://github.com/maximofn/langgraph_deepresearcher
Tau
I worked with the Alvea/Turing Dream team on a multi-agent tutoring system for secondary-school students. My contribution included planning, tools, context management and human-in-the-loop flows. It was used by real students in Spain and Colombia with teacher-validated course material.
https://app.taututor.com/
Horeca chatbot
I developed a voice-enabled agent for kitchen and dining-room workflows at CIVIR, including recipe and menu queries and updates. My work covered FastAPI APIs, web interfaces, retrieval integrations and deployment. Voice activity detection and buffered audio sent only real speech to transcription.
https://www.maximofn.com/projects/
Naviground
At Sener, I worked on perception for an autonomous ground vehicle: segmentation, classification, detection and depth estimation. I built datasets and optimised inference on NVIDIA AGX Orin with TensorRT, reducing latency and memory use by about 30% for roughly 2% accuracy loss.
https://www.maximofn.com/naviground
All to Markdown
Personal documentation-conversion tool. An LLM follows next-page links to traverse documentation, while MarkItDown converts the content into Markdown. The tool also supports individual URLs for web pages, PDFs, YouTube videos and Office documents.
https://www.maximofn.com/projects/
https://github.com/maximofn/all_to_markdown
Open cameras viewer
Personal local web interface for viewing multiple RTSP cameras with go2rtc. I packaged the stack in Docker and added a preferences service to preserve layouts, visible cameras and assignments across browser sessions. Includes predefined multi-camera layouts.
https://www.maximofn.com/projects/
https://github.com/maximofn/open-cameras-viewer
Subtify
Personal subtitle-generation tool using Python and Whisper. It creates subtitles in the selected language and differentiates speakers using subtitle colours. This project demonstrates audio-processing and AI-service integration.
https://www.maximofn.com/subtify
https://github.com/maximofn/subtify
maximofn.com
I built and operate my multilingual personal website using Astro and Vercel, with Cloudflare as a reverse proxy, WAF and media CDN. It presents technical articles, projects, datasets and talks. This is a portfolio example of web development and deployment.
https://www.maximofn.com/
18 public cases: AI agents, web, Computer Vision, Docker and datasets, with selected images, public links and my specific contribution to each project.