DEBUG: PAGE=domain, TITLE=NelsonHall Blog,ID=1469,TEMPLATE=blog
toggle expanded view
  • NelsonHall Blog

    We publish lots of information and analyst insights on our blogs. Here you can find the aggregated posts across all NelsonHall program blogs and much more.

  • Events & Webinars

    Keep up to date regarding some of the many upcoming events that NelsonHall participates in and also runs.

    Take the opportunity to join/attend in order to meet and discover live what makes NelsonHall a leading analyst firm in the industry.


Subscribe to blogs & alerts:

manage email alerts using the form below, in order to be notified via email whenever we publish new content:

Search research content:

Access our analyst expertise:

Only NelsonHall clients who are logged in have access to our analysts and advisors for their expert advice and opinion.

To find out more about how NelsonHall's analysts and sourcing advisors can assist you with your strategy and engagements, please contact our sales department here.

Tech Mahindra Populii: Addressing AI Training Hypergrowth


Populii is Tech Mahindra’s enterprise gig marketplace for data collection, user studies, and microtasks. NelsonHall recently spoke with Populii executives about the platform’s current capabilities and growth plans in support of the exploding AI data training market.

Human annotation and specialized microtasks

Populii is a gig community platform offering data collection of language, utterance, and accent samples, speech and audio, images and video. Populii clients can also run end-user device testing of hardware and software prototypes, collect healthtech PII and non-PII data, and conduct user behavioral studies. Projects can last several months, but only a few hours with microtasks such as annotation, metadata tagging, cleaning, and categorization of texts and NLP, ML and AI, audio and automatic speech recognition, and translation. Content types include computer vision, maps, video, images, 3D models, and various types of user-generated content and ads.

Microtasks also range from simple search, transcription, content quality rating and checks, to feedback classification and content moderation, to specialized tasks such as medical writing and piano roll reading. Populii provides consultants with extensive domain knowledge and specialized skills for these custom tasks. For example, for an American multinational technology corporation, since 2023, the Populii community  have analyzed, solved, and provided comprehensive reviews of math problems. Experts apply mathematical principles and methodologies to solve advanced math problems in elementary mathematics, algebra, statistics, geometry, and calculus. The global community members process, on average, 300k problems and ~1m annotations per month in ~14 different projects, achieving above 90% accuracy across projects.

Human cloud management

Tech Mahindra launched Populii in 2023 to help internal employees handle internal data training needs. It has since opened it to external workers and clients. The community currently has ~200k active users. The freelance workers support ~80 languages from ~100 geographies. For example, for a search engine client, Populii workers annotate images with the relevant objects of interest. Activities include creating content and enhancing data for problem database flow, segmentation flow, and domain flow in ~100 projects. For this account, Populii involved ~25k crowd workers in ~25 languages from Noida and Hyderabad, India; Lisbon, Portugal; Budapest, Hungary; and Malaysia. It delivered a 100% on-time completion for the client, a 25% reduction in AHT, and an 80.4% agreement rate achieved with a consistent quality achievement score above 95%.

Tech Mahindra recruits gig workers with targeted outreach using ads for flexible, part-time, and WAH jobs. For the math-solving projects, it ran social media campaigns targeting domain-specific communities. To reach further scale and acquire specialized skillsets and contributor profiles, Populii also partners with other gig platforms.

Populii employs unmanaged and managed crowd-operating models for three levels of client needs:

  • Unmanaged crowd, a low-cost model for collection, annotation, and curation requirements from the crowd pool with qualification criteria to perform tasks. It features metrics-driven performance monitoring, learning bites circulation, webinars, scorecards and stack ranks, and incentives. Here, the resources are completely scalable in a fully distributed crowd, and the client defines the price
  • Crowd managed is a medium-cost model for clients with 70-80% quality needs where Populii runs daily random manual sampling, error pattern analysis and implements quality monitoring dashboards and ranking. The focus for these projects is on selecting qualified gig workers and meeting the agreed SLAs
  • Fully managed crowd model is for quality needs above 90%, such as critical annotation or curation with SMEs on the platform and 100% commitment to timelines and quality. To ensure these targets, Populii employs quality monitoring dashboards, continuous monitoring, pop quizzes, spot checks, manual sampling, error pattern and gap analysis, regular training, knowledge assessments, and calibration sessions. Typically, Populii involves in-house SMEs for these projects to deliver above 95% quality and on-time completion. This model is closer to an FTE extension.

The platform allows pull and push of tasks to gig workers, including via auto-assign functionality. For the math-solving projects for the technology client, Populii applies a zero-disruption approach to transition where the freelance workers are taken on in a phased approach. The math problems are allocated based on past performance, with quality and eligibility status reviewed periodically. Populii deployed strong domain SMEs and an experienced crowdsourcing management team which performs validation and verification processes to onboard authorized experts. It also has a knowledge repository for continuous learning and upskilling.

Other features of Populii are freelancer qualification and certification, geography restrictions, user technical support, dynamic pricing, and automated payments to workers. For math-solving, the skills assessment includes eligibility criteria, proficiency tests, hands-on evaluation, and a final assessment, which takes up to a week. In the end, eligible workers are compensated for their assessment and training time.

AI dependent on humans

The current hypergrowth of GenAI is built with the large-scale data labeling work of platforms such as Populii, where gig workers enable datasets for efficient LLM training. Competitor examples include Wipro’s Topcoder, TELUS International AI Data Solutions, TaskUs’ TaskVerse, and Movate OnDemand. For many of these companies, the AI training work is the fastest growing service line.

Many of the leading high tech corporations have extensive experience with crowdsourcing, including with their own platforms. Tech Mahindra has set up Populii to couple the worker community acquisition, engagement, and recognition with the actual work execution tracks to easily direct gig workers to operate in the client environment. The company has several such implementations. As part of these engagements, Tech Mahindra still defines the minimum required skills, which guides the rate per task.

Populii leadership envisions the platform evolving to an end-to-end business process environment where complex processes such as claims management are broken down into microtasks for efficient handling by AI or humans. In the breakneck market for new product launches, Populii estimates it already delivers a 20-30% faster time to market at a lower cost. As a first step, Populii management aims to disrupt Tech Mahindra internal processes.

No comments yet.

Post a comment to this article: