ddv

Joined 13 years ago

Karma

304

Submissions

Member since

Jul 8

engineering and software development

View on Hacker News

Recent submissions

Comment · 2 years ago

Same here. Both copies I ordered were not working on arrival. I tried to switch the batteries but they were really stuck in there (can’t easily remove them without bending/breaking), eventually got them out and replaced them but one of the copies still didn't work. Reached out to the creator but he just said try swapping the batteries. I still think the idea of the book is great, just wish the execution and support were better.

Comment · 3 years ago

Location: Vancouver, Canada Remote: Yes Willing to relocate: No.

I'm an experienced machine learning engineer, software developer, and author of 2 patents in ML/AI. I'm an EGBC-licensed professional engineer and I have a BASc in Mechatronics Engineering and 10 years of overall experience in software development. I've developed and shipped end-to-end machine learning products and features across different platforms (web services, mobile apps, and embedded systems). I enjoy building practical machine learning systems from prototype to production that have a positive impact on people's lives.

Thanks for your time.

Technologies:

DL/ML: PyTorch, Tensorflow, Keras, data pipelines, ONNX, TFLite, Scikit-learn, OpenCV, OpenAI Api, Huggingface Transformers, Librosa, PyWavelets

Mobile ML: Swift, CoreML, iOS development

Data science: Pandas, NumPy, SciPy, Jupyter Notebook, Colab, Seaborn, Chart.js, Streamlit, Airflow, Dagster

Backend: Python, C++, Django, Flask, FastAPI, SQLite, PostgreSQL, REST, Docker

Embedded Software: C, Bare metal software development, TinyML, RTOS, Embedded Linux

Cloud: AWS, GCP, Linode

Website: https://dmitridevaz.com

Mojo Docs docs.modular.com

1 ddv 3 years ago 0 comments

Comment · 4 years ago

https://ciechanow.ski/gps/

https://ciechanow.ski/archives/

Comment · 4 years ago

Clear and concise write-up, thanks for posting. I especially appreciate the comparative graph at the end.

In practice, I tend to use "method 3" most of the time as I feel it has the best balance of complexity and practicality while still being able to deal with label noise.

Comment · 5 years ago

"Is there plan to support sparse NN influence on mobile devices, such as Arm CPUs?"

"Based on the amount of work for support, it's on our medium to long-term roadmap currently."