Umor Faruk Supto’s Vision: Building Bangladesh’s Place in the AI Training Economy
Meet the Chakri.ai co-founder and CTO, learn why he wants Bangladesh to contribute more than cheap labor to AI training, and see how Mercor and Outlier connect people to real model-improvement work.
By Chakri AI Research
When people talk about AI, they usually point to the model: the new capability, the benchmark score, the demo. But every useful model also depends on human judgment—people who write examples, catch mistakes, compare answers, and know when a response is subtly wrong. Umor Faruk Supto’s story connects a builder in Bangladesh with that quieter part of the AI economy: the work that teaches and evaluates models.
Supto’s public profile describes him as a computer science student at Rajshahi University of Engineering & Technology. His public project footprint is hands-on: his PyPI account lists Bangla OCR and Medunify, and a LinkedIn post says a VS Code extension he built passed 800 installs. He has also described Luminex Labs serving more than 25 clients across software projects. Those numbers are self-reported milestones, but they show a builder moving between software, users, and delivery—not only talking about AI from a distance.
Chakri.ai’s current About page identifies Supto as co-founder and CTO alongside CEO Amit Chandra Roy. That page describes Chakri.ai as an AI-powered hiring platform. This profile focuses on Supto’s individual builder story and the wider ambition he has shared: helping Bangladesh become a trusted source of talent for AI training and evaluation. It does not announce that Chakri.ai currently operates a Mercor- or Outlier-style contributor marketplace.
AI training is skilled work, not just clicking labels
A model-training project can ask a contributor to write a realistic question and a high-quality reference answer, compare two model responses, explain which one is more accurate, or test a system with a difficult edge case. Other projects need specialists to judge a legal analysis, a medical explanation, a software change, or the naturalness of Bangla phrasing. The work varies by project; the common thread is applying a clear standard and explaining the judgment.
Mercor and Outlier: two routes into the work
Mercor presents itself as a talent platform for AI projects and also works with enterprises on benchmarks, evaluation, and expert data. Its opportunities include project-specific roles and standing expert networks. Depending on the assignment, a contributor might create examples from professional practice, rate model outputs, or help evaluate an AI system against a domain standard. Some roles are open applications to a talent pool rather than an immediate paid project; Mercor says matching work can arrive later, so joining a network is not the same as receiving guaranteed hours.
Outlier, operated by Scale AI, describes its platform as connecting subject-matter experts with AI companies to improve generative models. Its work can include drafting examples, evaluating answers, and completing project-specific skill screenings. Outlier’s current contributor policy lists Bangladesh among supported countries, and a current Bengali freelance-writing opportunity includes Bangladesh in its eligible locations. Availability still depends on the particular project, qualification, and location rules.
| Mercor | Outlier | |
|---|---|---|
| What it is | Expert talent networks and project-based AI training, evaluation, and enterprise benchmark work | A contributor platform for human feedback and training tasks for generative AI |
| Examples of work | Write prompts and reference answers, rate model output, or evaluate a professional task | Create or review examples, compare responses, and complete skill-based project tasks |
| How projects start | Profile and assessment, then a role or network match; some networks may wait for a suitable project | Platform onboarding, identity checks, and project-specific qualification |
| Bangladesh access | Check each listing: eligibility and location can vary by role and client | Bangladesh appears in the current country policy and selected Bengali opportunities; each project can still set limits |
| Pay and continuity | Terms are listed per role; a network is not a promise of steady work | Rates and payment details are shown per project; weekly processing does not mean a fixed weekly income |
Other names in the contributor landscape
Mercor and Outlier are not the whole market, and the names below do not all offer the same kind of work. Their official pages describe different contributor routes—from general AI response evaluation to large-scale data collection and locale-specific search assessment. Treat this as a starting map, not a complete or ranked directory; open roles change often.
| Company or platform | What its official pages describe |
|---|---|
| DataAnnotation.tech | Human feedback and evaluation work, including reviewing model responses, writing, and coding tasks; applicants complete qualification steps. |
| CrowdGen by Appen | A contributor platform for Appen data projects, including AI data work; the company says its payment platform supports contributors in over 200 countries, but individual projects remain location-specific. |
| TELUS Digital AI Community | A job community with language-specific remote evaluator and data roles; availability depends on the locale and current opening. |
Do not read these platforms as salaried job boards with guaranteed hours. The opportunities are project-based. Listings change, projects can pause, and eligibility depends on the specific client and assignment. Read the live listing before doing unpaid assessments or sharing sensitive information. Confirm the rate, how time is counted, payment method available in Bangladesh, project duration, and any rules about confidential data.
What Bangladesh can contribute
The opportunity is larger than recruiting people to label data cheaply. Bangladesh has Bengali language expertise, educators, software engineers, clinicians, finance professionals, and people who understand local services and culture. Those perspectives can help test whether AI systems understand Bengali beyond word-for-word translation, follow local context, and handle domain work responsibly. A strong contributor is not interchangeable labor; their context and careful reasoning are part of the training signal.
To make that opportunity durable, the ecosystem needs clear task instructions, fair and visible rates, reliable payment, quality feedback, privacy safeguards, and pathways to specialist work. Universities and training groups can teach evaluation skills: write a rubric, support a rating with evidence, protect private information, and recognize uncertainty. Companies should make projects legible enough that contributors know what their work is for and how it will be judged.
Supto’s larger idea is compelling because it links a local builder’s work to a global shift in how AI gets made. Bangladesh can aim to contribute not only software and users, but also the language knowledge, professional standards, and human evaluation that make models more useful. That future will depend on how well platforms and local institutions value the people doing the training—not just on how many people they can recruit.
Sources and further reading
- Umor Faruk Supto’s personal website
- Umor Faruk Supto on LinkedIn
- Umor Faruk Supto on PyPI
- Chakri.ai: About and leadership
- Mercor: Experts and AI training work
- Mercor: Current opportunities
- Outlier: Contributor overview and FAQ
- Outlier: Supported countries and work policy
- Outlier: Current Bengali freelance writing opportunity
- DataAnnotation: Official contributor FAQ
- CrowdGen by Appen: Contributor platform
- TELUS Digital: AI Community opportunities
Specifications and prices can change. Check the linked pages before making a purchasing or deployment decision.