Dmitry Dumik

Handl - Turn paper documents into structured data

Handl converts any document into structured data. Powered by AI that is constantly improved by the data labeling platform.

Add a comment

Replies

Best
Ivan Uvarov
@dimadewinn Looks good! What do you mean by special knowledge from in-house team? What can your crowd annotate?
Dima Dewinn
@ivan_uvarov many thanks! Handl crowd is good for general tasks. You can assign tasks to groups of workers each of which is focused on a specific area. But if labeling requires the knowledge of Chinese or medical education, you may need your in-house team.
Shashkova Natalia
@dimadewinn @ivan_uvarov one soil one love
Dima Dewinn
Greetings Hunters, Dima from Handl here. I’m thrilled to introduce you to Handl, a tool to label and manage data for machine learning. On Handl, you will get your high-accuracy datasets with ease. We employ 25k qualified crowdworkers, who have labeled more than 6 million images, texts and sounds for tech companies and startups so far. Handl crowdworkers work remotely mostly from developing countries and get paid up to 3$/ho for their effort. If your labeling requires some special skills, you can invite your in-house team to do the job. We have a complete set of tools to cover data annotation needs: classes, bounding boxes, polygons, text input, and text segmentation, all easy to use and neat. Select and combine them the way you like. Above that, you can manage, share and safely store your datasets from here. Unlike MTurk and similar microtasking services, Handl stands for machine learning data labeling only. This allows us to acquire, train and qualify our crowd to perform labeling at the highest accuracy level on the market. Our consensus algorithm ensures quality by assigning the same task to a number of crowdworkers, until the proper accuracy is reached. Try Handl here — https://handl.ai On the occasion of the launch, Hunters get free annotations of up to 1,000 images or texts with the “PRODUCTHUNT” code. Follow the link — https://handl.ai/form We are happy to get your feedback and answer any questions. Cheers.
Tony Urban
Yes, yes, yes!!! We'd definitely use it
Dima Dewinn
@tony_urban looking forward to it!
Anna Gotta
Hi Dmitry, please provide more usecases on how your product can be used by different companies with different goals. I don't understand, if your product can be helpful for me.
Dima Dewinn
@anna_gotta we can't share details about our customers, but summarily more than 6 million annotations have been already done on Handl for companies as Nvidia, Nestle, Cherry Home, etc. Whatever they do with it)
Aaron O'Leary
Definitely going to try this out! Also I love the thumbnail!
Dima Dewinn
@aaronoleary pleasure to hear that!
Roman Tezikov
Nice idea! But why are you better than Mechanical Turk?
Dima Dewinn
Thanks @roman_tezikov. Unlike Amazon’s MTurk and similar microtasking services, Handl stands for machine learning data labeling only. This allows us to acquire, train and qualify our crowd to perform labeling at the highest accuracy level on the market. And we have a different internal workflow for data annotation — crowdworkers don't choose what tasks to perform. They just work properly and get paid based on the time spent and their accuracy reached.
Dmitry Tuzoff
Hi, quick question: are you planing some integrations with service marketplaces, like Freelancer or Upwork?
Dima Dewinn
@tuzoff many thanks! We do plan some nice integrations, we will announce them in detail soon.
Kate Khoshabova
Wish Golden Kitty for Handl's kitty! Meow! 🐾
Ravil Zaripov
Looks interesting. What is the difference between you and Figure8?
Dima Dewinn
@ravil_zaripov our pricing is much more accessible and its structure is different - we charge by hours instead of labeled data points. We go through all the husle with setting up the task and coming up with proper instructions in multiple languages for labelers by ourselves, and can provide assistance on machine learning-related projects beyond data annotation.
Andrew Yaroshenko
Looks really really interesting! Questions: 1) What is the maximum and the minimum data volume for labeling? 2) Can we buy ready "cats" (for example) vertical datasets? Is it going to be a marketplace for different verticals? 3) What other datasets verticals do you have?
Dima Dewinn
@goldenalf13 thanks Andrew! We can deal with any data volume and we have a flexile pricing for that. For now, we work with client's data only. We do not trade in any way.
Mikhail Larionov
So excited for this release. Well done @dmitry_matskevich ! What's your pricing?
Dmitry Matskevich
Thanks @mikhail_larionov! Pricing for our tools starts from $399/mo. Crowd workforce is charged separately, pricing depends on complexity of your annotation tasks. Learn more here — https://handl.ai/#howitworks
Andrey Klen
@dimadewinn Congrats on the launch, really dig the video. How does Handl handle data safety?
Dima Dewinn
@_klen thanks! We guarantee data security as we never show the whole data to a single worker. For example, when labeling IDs, we divide them into several fields so that one worker does not see the whole document. Later on, we put the fields together inside the client's contour.
Artem Gladkikh
Great job guys!
Mitya Sudakov
Quite excited! Do we pay fees when using the crowd onboard?
Dima Dewinn
@mitya_sudakov using our tools and labeling by the crowd are separate full-fledged options. Check out our pricings here: https://handl.ai/#howitworks
Anya Pozniak
I like the product. Great job, guys! Do you work with videos or just images & sounds?
Dima Dewinn
@anya_pozniak yes, we do. Usually we cut them in frames and label the resulted images. If you need to classify actions on videos, you can upload them in video clips.
Batyr Tanatarov
Thanks a lot, guys! A great product! So you'll be able to automate a large chunk of content managers job, right?
Dima Dewinn
@batyr not really, Handl stands for machine learning data annotation. Mechanical Turk or similar would better fit jobs related to content management.
Alexey Utkin
Guys, cool stuff, congrats! Will try this for our TV programmatic solution. I just wonder what languages do your labelers work with?
Dima Dewinn
Thanks @alexey_utkin! We work mostly with English but any common language can be used, if needed.
Alex Pustov
Cool video and nicely-done product. But who is your target audience?
Shashkova Natalia
@alex_pustov We work closely with machine learning teams from technology companies and startups helping them to deal with data preparation of any kind.
Alexey Melnichek
Do you have API channels for ongoing data flows? And if so, how do you charge for that.
Shashkova Natalia
@alexey_melnichek We do provide an API channel for recurring tasks. Text me at ns@handl.ai, let's see what we can offer you.
Alexey Petrov
Just tried it out. How fast can I get my dataset of up to 100,000 images labeled?
Shashkova Natalia
@alexeypetrov It really depends on the type of annotation and complexity of the task itself