Thread by @misterzedpk

Open Source Investigation – SurveyAuto and Dr Umar Saif

This research, based on open sources, tries to highlight issues and raise a citizen's legitimate questions/concerns around the use of a crowdsourced data collection app [Thread]

On 10th December 2019, a global press release was issued by SurveyAuto claiming they had received a (undisclosed) grant investment by the Bill & Melinda Gates Foundation to "monitor family planning product data".
archive.vn/vxwHV

The PR mentioned that the Foundation would rely on SurveyAuto’s platform for crowdsourced field data collected from clinics and pharmacies in Pakistan (basically field information which is appropriately geo-mapped through the app).

It was second engagement between the Foundation and Saif. Earlier in May 2018, Gates invited Saif at Seattle to learn more about his location-based data collecting apps which ‘improved immunisation’ and ‘assisted in crop yield estimation’. archive.vn/fbq4h

So, what exactly is ‘SurveyAuto’ and when did it emerge? Let’s examine chronologically.

Saif posted a job through LinkedIn where he mentioned SuveyAuto as "my startup in Lahore". Indeed, the same was mentioned in the company's PR that they have offices in Lahore (Pak), Amsterdam (Netherlands) and Boston (US).

Open source findings reveal that SurveyAuto BV was registered in Amsterdam on 21st September 2018 (KVK Number 72620196). Source: archive.vn/Bkwq0

KVK = Unique registration number alloted by Netherlands Chamber of Commerce.

SurveyAuto BV (Netherlands) has just "1 Beneficiary" (most likely Saif). Registered on address: Jan van Goyenkade 8, 1075 HP Amsterdam, Netherlands. Source: archive.vn/wip/6Tkm4

Another source says SurveyAuto BV has 11-50 employees. Source: archive.vn/bNUmb

On 30th October 2018, Saif is invited as Speaker at World Bank's Main Complex in Washington D.C. where SurveyAuto is among the projects he discussed in his (then) capacity as Chairman PITB. Source: archive.vn/aswg8

Few days later (2nd November 2018), Saif spoke at Harvard on same issues in (then) capacity of Chairman PITB. Interestingly, Harvard referred to SurveyAuto as "their platform" (viz Government of Punjab) meant to "inform policy in the govt". Source: archive.vn/dqf2u

After his moving-out from PITB, Saif was invited in personal capacity as a speaker at Startup Grind's Lahore Chapter (in partnership with Google) on 26th January 2019. This event's profile described Saif as "Founder and CEO of SurveyAuto". Source: archive.vn/yIG8l

Findings reveal SurveyAuto was first uploaded to Google Play on 10th April 2019. 'Mustansar Saeed' was identified as name of the app's developer/ publisher. Source: archive.vn/5M8qb

The app's description mentions the benefits of crowdsourced data collection and ominously includes the line: "Knowing what an electorate thinks can sway the course of elections". Source: archive.vn/vyWDa

Mustansar Saeed's LinkedIn profile reveals a strange anomaly. Apparently he was Software Engr at PITB from Nov 2015 till Aug 2018 but worked simultaneously at SurveyAuto a month before he moved over. Bizarre, is it not? Copy of LinkedIn CV: filebin.net/w2o584i8bv0svp…

Saeed has an impressive portfolio as he led software teams for multiple govt departments such as Education, Health, Overseas Pakistanis, Saaf Punjab and NH&MP, to name a few (LinkedIn info).

His developer account says he worked on Face Detection and Eye Scan apps also, officially. Source: archive.vn/PEtrB

The agricultural crop yield stat app for which Saif was praised by Gates, "Food @ Home", was a project by ITU Lahore. The attached process flow provides a framework to understand how SurveyAuto also functions (almost same workflow). Source: archive.vn/BVC1R

Coming back to SurveyAuto: From May to June 2019, the 'startup' completed a pilot phase project with World Bank. Source: archive.vn/10ajc

SurveyAuto recorded price and availability of 15 different commodities in 4 cities of South Punjab and Balochistan along the western route of CPEC. In this process, 400 wholesale/ retail shops were precisely geo-tagged via satellite by 20 field operators.

Creepy Part: SurveyAuto taught its AI system to analyse features like structure, geometry, colour and layout of shops through Machine Learning. Enabled app to automatically identify similar potential spots for geo-tagging (eye-opening, no wonder Gates was impressed).

Ongoing Project: 2019 Annual Status of Education Report (ASER) using SurveyAuto to collect data on household education in 9 districts using 540 field operators. Houses mapped via satellites with assistance of AI. Funding by UK’s DFID and UNICEF. Source: archive.vn/niBo3

Ongoing Project: Idaara-e-Taleem-e-Agahi (ITA) using SurveyAuto to digitise survey responses over 2-year period 'Advancing Action for Adolescent Girls'. Project ongoing since March 2018, implies Saif was running project during Chairmanship of PITB. Source: archive.vn/htQcS

SurveyAuto says its AI plots survey targets using Satellite Imagery, Call Data Records (CDR) Analysis and open street maps etc.

Have a look at dashboard metrics and insight.

Mustansar Saeed of SurveyAuto (formerly PITB) revealed (perhaps inadvertently) that one of the company's clients is "J-PAL" (Source: filebin.net/w2o584i8bv0svp…). This isn’t disclosed in any PR by SurveyAuto.

J-PAL is Abdul Latif Jameel Poverty Action Lab based in MIT (Saif's alma-mater). J-PAL's South Asia Centre is based in New Delhi, India and all partnerships/ research work with countries such as Pak are coordinated from here. Source: archive.vn/W13nT

Former Indian Administrative Service (IAS) officer-turned-scholar Iqbal Dhaliwal leads J-PAL's South Asia Office through liaison with MIT (US). Source: archive.vn/4kGsK

Question/Concern: According to Harvard, SurveyAuto was part of the Government of Punjab’s projects. Was it taken over by Dr Umar Saif discreetly or was it always his project which he was engaged in during Chairmanship of PITB? Needs to be ascertained.

Question/Concern: Mustansar Saeed who worked at PITB on critical and sensitive projects was apparently, according to his own LinkedIn portfolio, working on SurveyAuto in parallel. If it was Saif’s project, was Saeed being employed illegally?

Question/Concern: Examination of PITB/ ITU apps employing geo-tagging features suggests similar technology was adopted for SurveyAuto. If it is Saif’s project, were official resources being used for a private project?

Question/Concern: Details about World Bank’s May-June 2019 pricing survey which SurveyAuto undertook could not be found on World Bank’s website or any other open source mediums. What are the project details, if any?

Question/Concern: Why was the World Bank interested in geo-tagging 400+ shops along the very sensitive western route of CPEC?

Question/Concern: As per SurveyAuto’s own claim, Machine Learning enabled their AI to map facilities having similar features based on primary human-fed sources. What if this same tech is used to, maybe inadvertently, map sensitive national security installations?

Question/Concern: Are relevant Govt of Pakistan organs in Interior, Defence, IT & Telecom and Planning & Development ministries aware about this large-scale geo-mapping project carried out by SurveyAuto?

Question/Concern: SurveyAuto is registered in Netherlands outside Pakistan’s jurisdiction. Is it registered anywhere with FBR, SECP or MoI to ensure it comes under the purview of local laws?

Question/Concern: Did Dr Umar Saif misuse state resources for development of SurveyAuto? It’s evident he persuaded a skilled human resource like Mustansar Saeed to move over from PITB to his setup.

Question/Concern: Why is personally-identifiable household education, health and other data being mapped without the formation/ approval of a government-mandated taskforce? Can anyone setup data collection with private field operators and start mapping without clearance?

Question/Concern: Where are the back-end servers/ databases of SurveyAuto located? Is all data hosted within architectures in Pak's jurisdiction? Does the company have a CISO or risk mitigation strategy to prevent exploitation by hostile elements?

Question/Concern: Why did SurveyAuto mention the efficacy of digital surveys for electoral gains? Does the company intend to work along similar lines as Cambridge Analytica before the next elections?

Question/Concern: Were existing geo-mapped data entries of schools and health centres held by PITB/ ITU apps used to upgrade SurveyAuto’s own private database?

Question/Concern: Under whose watch or authority is Dr Umar Saif on the track of becoming the largest private custodian of critical public data? Will he use this as a leverage for contracts with the state in the future apart from NGOs/ INGOs?

Last Question/Concern: Is someone in the state noting all this?

The objective was to highlight issues and raise pertinent questions in public interest. Being a citizen, I exercised my due right. I have not, and will not, accuse any individual or entity of participating in anything detrimental to national security.

I leave my concerns/ open source findings for public record in the hopes that interested journalists, civil society rights groups or government quarters might want to probe further through their respective resources.

@threadreaderapp unroll