stevenminhhh

I just moved from a full-stack engineer position to AI researcher of an R&D lab in a software company.

Honestly, not much of a career change, but given the new position, I can be much more exposed to new technologies, get to try out new things, like IoT and Blockchain as well. Hopefully with the new knowledge I acquired, I can start my own business someday. Mostly since somewhere along the road of coding 8-to-5 as a developer, I want to create something for myself instead of building other's dream.

I think ClickUp can be a good alternative here.

It also have an Android app if that is one of your key requirements.

Thanks. I have been dealing with ton of headache from my projects since modifying PDFs can be very problematic. Rather than stitching up multiple libraries, I would rather suggest one platform to handle everything.

Will definitely keep this in mind until it meet my requirements. Is there a mailing list I can sign up for?

Do you have any plan in your roadmap to support different languages in the OCR feature? I'm specifically interested in recognize and processing PDF files written in Japanese and Korean.

I am also dealing with some clients that are struggling with processing handwriting in their document, but I guess it will be a little far fetched.

Unfortunately, the problem is the same with many compiled application out there, not just ML. My rule of thumb in using models from these public model zoo is to always use a separate container instance (or Colab) to check the source code.

Then either convert such model to some standardized format like ONNX, which might not be possible for some state-of-the-art models with custom ops afaik.

Or you could just load the torch model first, then load the model state dict separately. Avoid using command like torch.load() since it uses pickle module implicitly, since it is possible to construct malicious pickle data which will execute arbitrary code during unpickling.

I think the best way to learn is to give yourself a meaningful project to start with, that can showcase the power of containerized architecture. Some suggestion would be to launch a LAMP or MERN stack using containers, and deploy it on AWS ECS for example. Also given the recent update to the Docker Desktop terms, I think it's better to switch to Linux and use the Docker Engine through CLI instead of Docker Desktop if possible

I think there is a fine line between iteratively developing your product as a whole and leaning it completely toward your biggest customers. It might not be ideal to turn your product development team into a software solution team that trying to meet every customers' needs.

These are in general good ideas that we'd like to build anyway in time.

Meaning these features are already somewhere in your product roadmap. How about just share your future development plan and give them access to professional customer support? Anything further than that like shifting development effort to their needs completely just feel wrong in a product point of view. Sell your most valuable features, not your dev team.

Are there alternatives means of partnering with customers that would give us the resources to develop faster without loosing our focus on the overall product and commitment to quality over deadlines?

This might be highly unrelated due to different business model, but for my previous startup I stuck in the same boat for a while. Eventually, I made part of my product open-source, build a community of fans that love the product. Afterward, the development support and contribution from the community is enough to keep the product going.

I don't think any Russian developers deserve this at all. Kinda double-standard in my opinion

Having some experience in software solution related to these platform, I think too many companies are using these AI-powered recruitment platform in a wrong way. It should be a mean to filter out extreme outliers in a high volume staffing scenario, not as a convenient and automate analysis for employers to make hiring decision on.

AI is built on hundreds of thousands of data points, to accurately utilize it you have to depend on people to input the right data, in this case job seekers accurately listing abilities, skills or nuances of their experience within a specific industry. Given the slight idea of where these platform get their data, the AI model should give a yes for an applicant that work well as cog in the machine, which might worked well for most companies, but you might be missing out on the anomaly data/applicant make your startup the next unicorn.

Then again, humans are good at doing human things.