CIOInsider India Magazine
.png){kind=logo}
Sampath Herga
CTO & Founder
Taskmonk Technology, headquartered in Bengaluru, provides AI / ML to companies with a data annotation/labeling platform to build data together with annotation partners of their choice. The idea for Taskmonk came about from when the founders Chetan and Sampath got together to resolve data-related problems they were having in their own companies. More so, the founder will explain how they provided the services on a pay per task model, no setup costs, and no platform fees.
In conversation with Sampath Herga, CTO & Founder, Taskmonk Technology
AI and ML are gaining traction as they are touted to give businesses the edge which is necessary to improve efficiency and increase productivity. Please give us an insight into how Taskmonk has grown and adapted to changes in this domain?
A 2017 PwC report on the world economy states that by 2030 AI’s economic impact will be in the trillions of dollars. Deployment of AI is impacting all industries, hence companies need to embrace AI to adapt and survive in this new environment. To make AI successful in the real world, it needs human wisdom, enterprise grade quality, and scale. Human wisdom is delivered by large teams converting raw data to intelligence by providing tags, annotations, categorizations, etc. Due to
the ubiquitous use of AI, the number of use cases on which AI is applied is also very large. To accommodate the large number of use cases across data types such as text, image, audio and video, Taskmonk is built on a flexible and configurable architecture. Cloud based deployment allows for on demand scaling and payment on a per-task basis for our customers. In 6 months we have grown from delivering 500,000 tasks per month to 3 million on the platform across 3 customers and 20 different processes.
Our architecture facilitates the designing of top screens and workflows for text image, video and audio data process
Looking at the current trend, businesses are more inclined towards adopting AI and ML into their systems as the need for analyzing the big data needs efficient machines to do the job. In what ways does Taskmonk’s service assist the companies in achieving their business goals?
Companies today need to develop and deploy AI applications for several business processes across the organization. Generating labeled annotated datasets is a key constraint for developing real-world AI applications. Taskmonk empowers businesses to build high-quality datasets for AI models enterprise-scale with expert teams. With Taskmonk organizations can control, manage and optimize supply of annotated data to its AI models. API’s and allocation algorithms help in task management. Proprietary algorithms, embedded AI, UI scripting engine enhances productivity and quality providing teams the opportunity to generate larger datasets same budget. Also, Taskmonk works on a pay per task basis, for all data types
(Text, Image, Video, and Audio) on a single platform. Integration to a variety of human wisdom partners is permitted: inhouse teams, vendors, Taskmonk’s partners; hybrid teams can also be built. 3 million tasks per month are delivered on the platform across 3 customers and 20 different processes.
Kindly share a success story about one of your implementations that has earned a significant reputation for the organization.
A large retail enterprise was building an AI model to perform entity matching. For the training of their AI model, they required annotated datasets from their vendors. We worked with the annotation team and helped them incorporate ground insights from team members. Using field validation rules at the UI level though our UI scripting module and development of smart data sampling procedures we were able to deliver data accuracy at 99 percent. The annotation team also saw a 35 percent improvement in productivity (tasks per person per unit time) when the affinity model for data allocation was switched on. Affinity allocation enables homogeneous flow of tasks to the analysts, therefore, increasing familiarity, thus improving speed and accuracy. We currently deliver 1.2 million tasks per month for this customer.
Tell us about the future blueprint of the company for the days ahead.
We are very strong in the e-commerce vertical, with solid knowledge on most use cases. Modules for computer vision such as key point annotation, bounding boxes, polygons, polylines, semantic segmentation, and instance segmentation are ready. Additionally, our 3D point cloud annotation that helps to build 3D perception using LIDAR, camera and RADAR will be released next month. NER and POS capabilities are present for training NLP models. We are putting most of our R&D budget on developing automation algorithms, and we are also currently exploring the synthetic data literature as well.
