BinaMora | AI films and software by Eitan Ker

Ideas built with AI

Watch Natural Fear

Featured AI Film

NATURAL FEAR

An AI short film by Eitan Ker

NATURAL FEAR is an AI short film by Eitan Ker, made with Higgsfield. Around a secret round table, the Devil, the Reaper, an alien, an AI machine and a man debate how to bring humanity to its knees. You can watch the film on Higgsfield or on YouTube.

Curious how to make a film with AI? NATURAL FEAR is one worked example. Eitan Ker made it with Higgsfield, a platform for AI video creation, and the finished film is a good way to see what AI filmmaking can look like in practice.

BinaMora is the studio and site of Eitan Ker, an engineer turned filmmaker. Here he makes AI films such as NATURAL FEAR and builds AI software, including AI agents like Travel Assistant and AI Tutor, an AI tutoring project for university students.

AI Software

Earlier Projects

Parking Detector camera feed: occupied parking spots boxed in red by the YOLOv5 model, free spots boxed in green

AI · Computer Vision · 2023

Parking Detector

An early computer-vision project. A Python application watches a live IP-camera feed of a parking lot and reports, in real time, which spots are taken and how many are free. A pretrained YOLOv5 model detects parked cars, and a Flask web app serves the annotated stream and the free-space count, with spot data stored in MongoDB.

  • Parking positions are learned, not hand-drawn: OpenCV background subtraction and an object tracker record where each car stops inside a marked area.
  • Free spots are chosen by solving a maximum independent set of non-overlapping positions with dynamic programming.
  • Detection and position learning run on separate threads alongside the web server.
  • Python
  • YOLOv5
  • PyTorch
  • OpenCV
  • Flask
  • MongoDB
View on GitHub
Travel Assistant chat interface showing a traveler profile, an Iceland trip answer, data-source tags and supervisor PASS badges

AI Engineering · 2026

Travel Assistant

A travel planner built on Claude as a multi-agent pipeline rather than a single prompt. A travel agent with 10 tools chooses between live APIs (OpenWeatherMap, RestCountries, OpenTripMap, Frankfurter), web search and a local knowledge base, while three supervisor agents check the request, the fetched data and the final answer.

  • Supervisors return PASS, REFINE or CLARIFY; on CLARIFY the pipeline stops and asks the user (e.g. which Paris) before any tool runs.
  • Context is managed explicitly: a traveler profile is extracted from every message, tool results are cached per session, and long histories are compacted into a structured summary.
  • RAG over 322 WikiVoyage and Wikipedia documents embedded with Voyage AI; tool calls run in parallel.
  • Claude API
  • TypeScript
  • Node.js
  • Express
  • React
  • Voyage AI
View on GitHub

Get in Touch

Contact

Drop us a line.

Required