MillionX Bangladesh · Proposal to the University Grants Commission · Draft of 27 August 2026

The system, on one page

Three views of the same proposal. The first shows how a learner moves through it and where they land on the AI5K ladder. The second shows the two flywheels that keep the curriculum current. The third shows what each intervention delivers. Both map to the Commission’s five strategic priorities.

AI5K — one system, one verified targetCohorts move along the horizontal axis. Enablers and AI5K do not — they run underneath the whole width. $5,000 a month sustained and verified is the top tier of a published ladder, not a minimum every participant reaches.EntryYear 1–2Year 3Final yearGraduation+6 / +12 monthsTARGET OUTCOMEProgramme A — AI5K FoundationFirst and second year. Foundational digital literacy,reasoning, conduct. 10X Minds entry plus build-a-thontrack. Ends at tier A0–A1.UNDER WAY · Method proven at 1,000+ (BRAC, Jun 2026)and 500+ (Build-A-Thon, Jan 2026). Programme asscoped not yet run.ContinuationCapability appliedinside the discipline.Institution-led.Programme B — AI5K ProfessionalFinal year and graduates. Applied AI practice,portfolio over transcript, 3–6 month internship.Targets the credentialed-expert segment at$60–250/hr.UNDER WAY · Curriculum built and mapped; DIUimplementation under way. Internship pipeline notyet run at scale.In workRetention measured at 6and 12 months.NOT YET RUN · Nooutcome data yet.Baselines set in thepilot.Unemployed and underemployed graduatesEnter laterally at Programme B. Already hold thedegree; lack applied capability and a bridge intowork.DELIVERED · Cohort evidenced: BBS Labour ForceSurvey 2024 — 885,000 unemployed graduates, 13.5%rate. QS 2025 ranks Bangladesh 67th of 81 onemployer alignment.Lateral entry — no pipeline. This cohort is the fastest to convert and the largesttoday: roughly 885,000 unemployed graduates nationally.Programme C — AI5K AccessRural youth and school leavers. Upazila weekend hubs,Bangla-language tools. Annotation and labelling asthe entry — open access, $10–15/hr, region-tiered.NOT YET RUN · Not yet run. Pilot upazila to bedesignated by the Commission.Parallel track. Feeds tertiary entry or direct earning— it does not have to route through a degree.ZONE 1Who moves throughThree populations, threeentry points. Only thefirst is a four-yearpipeline.A0 · first dollarearnedProfile live, payment railtested. The step mostprogrammes never reach.A1 – A2 · $250 →$850$850 is parity with theDhaka freelancer average —today's ceiling.A3 · $1,500Top of the reported band.Replaces a graduatesalary.A4 · $3,000Specialist positioning;usually the point ofjoining a delivery pod.A5 · $5,000sustainedSix months, verified viaformal remittance. Aminority reach this.ZONE 2The AI5K ladderOne target across everyprogramme. Published infull, including how manynever leave A0.Curriculum — three AI-native tracks, already OBE-mappedProduct Engineering · Analysis · Organizational Culture and Leadership. Mapped to UGC, BNQF, CLO/PLO and Bloom’s, so they enter through existing approval pathways. Deliveredas stackable microcredentials.DELIVERED · Three curricula built and mapped to UGC, BNQF, OBE, CLO/PLO and Bloom’s. Implementation under way at Daffodil International University through microcredentialsand formal programmes.UGC Priority 3 · 2030 target: OBE and AI-basededucation in all universitiesFaculty — train-the-trainer, Tier 1 → 2 → 3Tier 1: MXB trains faculty at pilot institutions. Tier 2: those faculty teach their own students. Tier 3: faculty train faculty; the Commission governs scaling. Certified,not briefed.UNDER WAY · Faculty-facing curriculum built (AI-Native Organizational Culture and Leadership). Tier 1 has not yet run — it begins at the pilot institutions.UGC Priority 4 · delivered through existing HEATchannelsAssessment — classify, verify, recordAI-use classification in every course outline · at least one secured sitting verifying unaided attainment · the Capability Record as a transcript supplement. Optional;Section 8.NEW IN THIS DRAFT · New in this draft, not discussed on 25 August. Optional; removable without affecting the rest of the proposal.UGC Priority 1 · 2030 target: Internship andEmployment Monitoring SystemAI5K governance — verify, read the market, credential, referEarnings Verification (every tier claim checked against ERQ and Form C records) · Market Signal Unit (quarterly demand read) · Credential Authority (microcredentials mappedto BNQF) · Pod and Referral Desk.NEW IN THIS DRAFT · New in this draft. No participant has reached any tier — the framework is proposed, not evidenced. Every figure is published data or arithmetic on it.UGC Priorities 1, 2 and 3 · 2030 target: raisingICT exportsZONE 3What enables itFull width by definition.These do not sit at astage — they hold upevery stage.EmploymentPlacement and retention at 6 and 12 monthsNo outcome data yet · baseline in pilotChannel 1 — global marketplacesAnnotation $10–15/hr (Appen, Toloka, CrowdGen) · RLHF$18–55 · coding eval $65–110 · domain expert $60–250(Mercor, Surge, Outlier) · AI integration and agenticbuild $50–200 (Upwork, Toptal, Braintrust)Upwork AI skills +109% YoY · integration +178% ·annotation +154%The wall: payout coverageThe highest-paying expert platforms exclude Bangladeshon payment-processor coverage, not capability. Apayments gap, not a skills gapThe 29 July 2026 BB framework is the fixChannel 2 — NRB ConnectDiaspora-vouched SMB and enterprise work to deliverypods. No commission, no country filter on the rateProposed · 12–13m abroad, $35.5bn remitted FY26National arithmeticOne participant at A5 = $60,000 a year in verified FX.12,077 doubles national ICT service exports; 83,333reaches the $5bn targetICT service exports were $724.6m in FY25Skills and microcredentialsAssessed at entry and exit, all three components1,500+ already through structured trainingResearch papers and patentsProgramme F — diaspora co-supervision, Bangladeshi firstauthorsProposed · no delivery to dateResearch is on a different track. It is a postgraduate andfaculty output, not an undergraduate one — it branches atgraduation and does not come from Programmes A, B or C.ZONE 4THE EVIDENCE UNDERNEATH · DELIVERED WORK, NOT PROPOSED WORKJan 2026Genesis AI ConferenceUniversity of Asia PacificVCs of UAP and IUB, Pro-VC of BUET, BRAC CSEchair, UIU and WUST leadership; BRAC, ACI,Square Toiletries, Link3, Dhaka TribuneProves the convening problem is solvedJan 2026National AI Build-A-ThonDaffodil Plaza500+ participants across education,agriculture, healthcare, e-commerce andfintech, with BASIS Students Forum, DaffodilGroup, Varendra and MetropolitanProves the method works acrossinstitutions, not inside oneJan 2026MillionX Launch WeekendTwo days160+ young innovators, industry leaders,academics and policymakersProves sustained convening, not a singleeventJun 2026BRAC University journeyExtended practical track1,000+ AI builders prepared and engaged. Thelargest single cohort taken throughstructured training to dateProves depth inside one institutionOngoingThree AI-native curriculaImplementation at DIUMapped to UGC, BNQF, OBE, CLO/PLO andBloom’s; delivered as microcredentials,university courses and formal programmeintegrationProves entry through existing approvalpathwaysDec 2026International AI BuildersCongressDIU Ashulia, 26 DecemberSix-month builder journey; target of 5,000trained. A step up from the 1,000+ largestcohort to date, and we say soUnder way — not yet evidenceEVIDENCE SCALEDelivered · dated, numbered, substantiableUnder way · built or running, outcome not yet inNot yet run · proposed, no evidence claimedNew in this draft · removablePriority 1Employability, demand-driven educationPriority 2University–industry collaborationPriority 3Digital transformation and AIPriority 4Faculty developmentPriority 5Research and internationalisationMillionX Bangladesh · National Action Plan for Higher Education Transformation 2026–2030 · Programme F, Section 6 (earning channels) and Section 8 (verification) are new in the current draft and can be removed without affecting the rest.

Why these axes rather than a simple grid

The obvious version of this diagram puts your five items down one side and the four targets across the top. It fails, because the five items are not the same kind of thing. First-year students, final-year students and unemployed graduates are populations that move through time. Curriculum update and faculty development are enablers that do not move at all — they sit underneath every stage at once. Placing both on the same axis implies faculty development is a stage a student passes through, which is not true and which makes the diagram harder to read the longer you look at it.

So the journey view puts time on the horizontal axis and role on the vertical: who moves, what accumulates, what enables. Enablers are drawn as full-width foundation bars, which is both accurate and rhetorically useful — it shows the Commission that curriculum and faculty are the load-bearing elements rather than two programmes among five.

The one thing to check before this is shownResearch output does not come from the undergraduate pipeline. It branches at graduation and runs through Programme F, postgraduate co-supervision and faculty. Any diagram that puts “research papers” in the same row as first-year students is claiming something the proposal does not support, and a Vice-Chancellor will notice immediately.

The evidence layer

Every element now carries its own proof, on a four-point scale: delivered (dated, numbered, substantiable), under way (built or running, outcome not yet in), not yet run (proposed, no evidence claimed), and new in this draft (removable). Zone 4 sets out the dated track record underneath, and each card states what that item actually proves rather than just what it was.

Why marking the gaps is the stronger moveSix of the twelve elements are not yet delivered, and the diagram says so in grey. A Vice-Chancellor reading a diagram where everything is claimed equally discounts all of it. A diagram that distinguishes 1,000 builders at BRAC from a rural programme that has not started earns the parts that are real. The delivered items are the ones the Commission cannot get elsewhere; hiding them among proposals wastes them.

Two further points the layout makes without argument. Unemployed and underemployed graduates enter laterally, not at the start — they are the fastest cohort to convert and the largest today, and drawing them as a side entry at final year says so. And rural youth run on a parallel track that can feed tertiary entry or direct earning, which is the honest depiction: Programme C does not require a degree to pay off.

Using these

Both views are vector. Print this page to A4 landscape for a hard copy of each, or use the standalone SVG files, which drop into Word, PowerPoint or the proposal itself at any size without loss. The journey view is the one to put in front of the Chairman; the matrix is the one to leave behind with a working team.

MillionX Bangladesh — system diagram for the proposal to the University Grants Commission of Bangladesh
Draft of 27 August 2026 · Not for circulation · Programme F and the assessment layer are new in the current draft and can be removed without affecting the rest.