Electrical Design in Lighting:

• Preciosa Signature - Milano Design Week 2023, 2024 Signature design concept development, electrical design, prototyping, technology, drawings, preproduction, installation. Working in young flexible team, realising ambitious ideas in limited time.
• Great Ribbon - Hong Kong: novy projekt
• Interactive Light - Cayman Islands: novy projekt
• Electrician craft: Wiring, soldering, switchboard installation, blueprint reading/drawing, safety awareness, valid certificate 250/2021 (czech electrical eligibility decree).

Embedded systems and IoT:

• Home Assistant Automation: Automation of lights, speakers, electricity consumption measuring etc. with Home Assistant platform, running on Docker. Connected with additional platforms such as Node-Red....
• DMX Led modules: Various LED module designs with color changing and dimming capability onboard, using DMX 512 transceiver ICs.
• Softboxes: Two battery powered softboxes using tuneable white LEDs and controlled by ESP32. Using machine learning to fit CCT of the led to the surrounding light temperature.
• Skidsteer Toy: Toy project of skidsteer using 3d printing and Arduino programming.

• Rohlík Orders Forecasting Challenge Kaggle challenge of forecasting orders for Rohlík company using data science and machine learning. This embedded notebook employs LSTM neural nets to forecast orders with apx. 5% MAPE. The second notebook utilizing XGBoost method reached even better score
• Youtube Playlist Diarization: This tool performs video transcription and diarization (identification and timestamping of individual speakers within a transcript) on Youtube playlists.
• Project Time Tracker: Time tracking app, tracks projects you work on and keeps the effort neatly organized for each month in an excel spreadsheet, which it produces and formats.
• Audio dataset Creator and Augmentator: Records audio samples on microphone and helps you to produce and format audio dataset for ML.

• Sound Event Detection System: Machine learning system that can be adapted for tasks like sound detection, audio classification, keyword spotting etc. There is implemented a trigger system that controls ligting over Artnet communication. Github repo.
• Light Chromaticity Regression Model: Non linear regression model , realtime prediction a light temperature (CCT) based on input data from RGB photodiodes
• Keyword spotting tasks: Triggers an action based on a spoken word or series of words. Applying various feature extraction methods such as Mel Spectrograms and MFCCs, training machine learning models with Tensorflow, Pytroch and running inference.
• Gesture Light Control: Control lighting using hand gestures captured via webcam. The webcam feed is processed using Google Mediapipe, which employs an convolutional neural network for image classification. The recognized gestures then trigger an Art-Net signal, which is sent to DMX lighting control software.

• Stable Diffusion LORAs: Training Low Ranking Adaptors - LORAs (partial neural nets) of various diffusion models based on Stable Diffusion. Using platforms like Kohya, Automatic111 and training on Cloud GPU clusters.
• Supervised Fine-Tuning opensource LLMs: Training LLMs on custom data is greatly simplified with the help of the Hugging Face Transformers platform. I worked with models like Mistral7b, Llama2, Phi-2, trained on existing datasets or custom podcast, chat, personal data etc. Low GPU VRAM can be compensated for by 4-bit quantization and LORA training.
• RAG retrieval from Standards: Based on natural language question query retrieves an answer with specific technical details from company electrical standards. Works in Czech and English languages. Using Langchain and Chroma or FAISS vector database, powered by opensource llms and embedding models such as Mistral 7B and Sentence Transformers.
• Intention Detection from Speech: This app uses audio transformer (ASR) models to percieve an intention from spoken language and trigger corresponding action. The audio input query is recorded and processed in real-time. Uses finetuned Whisper model to classify the query.

• Dynamic Ligting Visualisations: Visualisation of Milano 2024 lighting installation reacting to 3d sound. Virtual installation reacts to real dmx512 values and accurately represents lighting behaviour. Visualisation can be rendered realistic or real-time with more stylised look.
• Virtual Environment Scenes: Producing a virtual environment, which can be entered, walked in, simulated, shot in and configured by the visitor.
• Render Camera Shots: Render test shots simulating real world cameras to visualise a product or scene.

• Dynamic, Interactive Light Design: Designing, patching, mapping the fixtures and producing blueprint. Designing lighting scenes in Madrix 5 and Marix Script.
• Lighting Design: Light source deployment and customisation for specific component lightup. Optics and diffusor combination for reaching the best lighting effect.
• Healthy Lighting: I'm interested in using dimmable, human-centric LED lights that follow natural circadial rhythms. Deploying LEDs with variable CCT, using tunable white modules, 2-wire dim-to-warm strips, deploying a custom ML system to regulate light chromaticity.

• Rhino-Grasshopper Building Structures: Parametrically defined structures, created in Grasshopper, used for Unreal Engine scenesor 3D printing and prototyping. This approach leverages the ability to make complex design changes through several input parameters.
• Rhino-Grashopper Data Engineering: Processing data from already created assembly model in Rhino, Grasshopper pulls out many valuable data. After cleaning, sorting and preprocessing, these data can be used for 3d mapping and patching of LED lights for example.
• Inventor: Used for parts and assemblies creation. Mainly led driver assembly boards, led components and detailed installation drawings. I'm familiar with Vault Pro pipeline.
• Blender: I'm familiar with Blender workflow, using mainly for UV unwrapping and basic 3D work.

• Davinci Resolve: Processing and postprocessing rendered shots, music and audio.
• FL Studio: Past experience in music production, mixing, and mastering, along with sound design using additive and wavetable synthesizers, and resampling techniques.
• Photoshop/Gimp: Photo and image editing, processing and combining. Producing and editing shader textures.