Defra

Defra: Artifical Intelligence: Discovery Alpha Beta

Incomplete applications

5
Incomplete applications
2 SME, 3 large

Completed applications

18
Completed applications
8 SME, 10 large
Important dates
Opportunity attribute name Opportunity attribute value
Published Thursday 17 May 2018
Deadline for asking questions Thursday 24 May 2018 at 11:59pm GMT
Closing date for applications Thursday 31 May 2018 at 11:59pm GMT

Overview

Overview
Opportunity attribute name Opportunity attribute value
Summary of the work To develop systems that use data sets in conjunction with capabilities such as AI, Machine Learning and Data Science within the inspection regimes at the UK border and in the maritime environments.
Latest start date Monday 9 July 2018
Expected contract length Up to 2 years. Discovery 8 weeks.
Location South East England
Organisation the work is for Defra
Budget range

About the work

About the work
Opportunity attribute name Opportunity attribute value
Why the work is being done Outcome: An IT system or systems which applies optimum use of AI to ensure the import of animals, plants and associated products are healthy and minimises cost and frequency of maritime fishing inspections in the first instance, expanding those systems into other areas as appropriate
Problem to be solved To apply AI to improve Defra operations, initially for Border and Maritime inspections including:
To better target consignments/catches that are likely to be non compliant through the use of various data sets and technology capabilities: Manual risk assessment is labour intensive and based on guidelines and factors such as country of origin, which lead to less effective inspections: Current inspection outcomes are not continuously incorporated into the process to enable improved detection rates. There is potential for further work beyond these areas
Who the users are and what they need to do For Intelligent Inspections at the Border and in the Maritime Environment, some of the users and responsibilities are:
1) As a Port Health Officer, I want to be told which inspections to carry out based on risk of non-compliance, so that I can efficiently locate non-compliant consignments.
2) As a Risk Analyst, I want my definition of risk factors to be guided by intelligent insight, so that inspection guidelines can be adapted dynamically to improve detection rates.
Early market engagement
Any work that’s already been done Defra has engaged with an external organisation to provide a high level review of analytics and Artificial Intelligence at Defra. Several potential opportunities to apply Data Science, Machine Learning and AI to Defra's work have been identified. Defra is now working with internal stakeholders to better understand and capture scope and user stories.
Existing team SMEs and managers from the APHA and MMO organisations within Defra.
SMEs and managers from the Food Standards Agency.
Staff from Defra's established Data Transformation Programme - Senior Managers, Delivery Manager, Engagement Lead, Business and Data Analysts, and Data Architects
Current phase Discovery

Work setup

Work setup
Opportunity attribute name Opportunity attribute value
Address where the work will take place Weybridge, London, with possible visits to York and Carlisle (APHA sites) and ports of entry, e.g. Felixstowe.
Working arrangements On site in Defra offices for majority of time with some scope for remote working upon agreement and adhering to Defra's various processes and policies i.e. security, data protection etc.
Security clearance Disclosure Scotland but may be some requirement for SC clearance

Additional information

Additional information
Opportunity attribute name Opportunity attribute value
Additional terms and conditions

Skills and experience

Buyers will use the essential and nice-to-have skills and experience to help them evaluate suppliers’ technical competence.

Skills and experience
Opportunity attribute name Opportunity attribute value
Essential skills and experience
  • "1. Evidence experience of developing analytics solutions to provide better insight to deliver measurable business outcomes using large, disparate and/or complex data sources. 10
  • Evidence of deep knowledge and expertise of using artificial intelligence and predictive modelling solutions and the identification and ongoing maintenance of data quality. 10
  • Evidence experience of rapid development of proofs of concept (PoCs) using agile methodology that are robust enough to base future decisions on and which conform to clients' environment expectations. 10
  • Experience of evolving PoCs through to production systems and scaling accordingly. 8
  • Experience of delivering solutions to front-line users in critical operational areas that are easy to implement. 8
  • Evidence by example of proven capability to deliver data science expertise into projects and share learning. 8
  • Evidence by example of using open source components to deliver the solution. 6
Nice-to-have skills and experience
  • Expertise in Agile working, delivering proofs of concept, and sharing learning and knowledge. 5
  • Familiarity with consignment tracking systems and/or health certification systems. 5
  • Evidence by example using open source components to deliver solutions. 6

How suppliers will be evaluated

How suppliers will be evaluated
Opportunity attribute name Opportunity attribute value
How many suppliers to evaluate 3
Proposal criteria
  • How the approach meets Defra requirements. 10
  • Estimated timeframes to onboard.4
  • How risks and dependencies are identified and proposed approaches to manage them 4
  • Team structure, capacity and capability 4
  • Value for money. 8
Cultural fit criteria
  • Work as a team with our organisation and other suppliers
  • Be committed to Agile working
  • Be collaborative and transparent when making decisions
  • Take responsibility for their work
  • Share knowledge and experience with other team members
  • Challenge the status quo
Payment approach Time and materials
Assessment methods Written proposal
Evaluation weighting

Technical competence

60%

Cultural fit

10%

Price

30%

Questions asked by suppliers

Questions asked by suppliers
Supplier question Buyer answer
1. 'Evidence by example using open source components to deliver solutions' appears as both an Essential and a Nice to Have. I would be grateful if you could clarify This criteria appears twice by mistake (with very slightly different wording) It should be a 'Nice to Have' only:-

Evidence by example of using open source components to deliver the solution. 6
2. Is this work for 'Discovery' only? Or you would select the same supplier for Discovery, Alpha & Beta phases? Completion of Discovery is the immediate need with an option to continue into the other phases
3. Do you have an in-house 'Data Science' team No
4. I am afraid you have already defined the 'solution' to use AI, ML & DS - but that might not be needed or needed in a less scope to fulfill the project goals based on User research. So, is the idea of p[roject to implement the AI, ML, DS capabilities, come what may? The PoC and beyond are intended to prove the business value of using ML, AI and DS technologies, so we would expect the solutions to demonstrate this.
5. What is the Budget of this 'Discovery' work?

Please could you provide the budget estimate that you have allocated to this project
Current estimate for this project is around £1 million. This project is one aspect of a wider project for which funds are available.
6. Can you please confirm this is contract is outside of IR35? We are tendering for a service, IR35 is not relevant.
7. Can you bring to life the data sets that exists already and how they will be accessed? We are working to identify datasets of interest. For Intelligent Inspections at the Border, data from the TRACES system

(https://ec.europa.eu/food/animals/traces_en)

is likely to be used.
8. Do you expect the successful supplier to bring an understanding of the maritime issues, or will these come from the team in DEFRA? No. SMEs from relevant Defra areas will have input to requirements.
9. Is there an existing supplier working/advising with Defra in this area? Yes. We have received advice from a supplier following a period of consultation.
10. You mention that there have been manual versions of approach used previously, are there conclusions of this work that we can build into the proof of concept? Our SMEs can bring knowledge of the approaches to identifying consignments and vessels for inspections. However, part of the challenge here is to use the data to identify patterns of consignment and fishing non-compliance rather than simply reproduce human-defined rules.
11. Can Defra please share the name of the organisation that performed the high level review of current analytics and Artificial Intelligence? ASI Data Science
12. Please could you confirm if the authority intends to let a contract for the full two years or just the eight weeks of discovery? The contract will be let for 2 years to enable any ‘Statements of Work’ required within that period
13. Please could you clarify what is meant by “delivering solutions to front-line users in critical operational areas that are easy to implement”? Are you looking for evidence of: suppliers’ ability to develop solutions that are easy to deploy; evidence that we can develop solutions that can be deployed into challenging environments, or evidence that we can develop solutions that can be used by people working in challenging conditions? The end goal is to provide information to users at the ‘sharp end’ of inspection activities to better target consignments and vessels in their everyday operations. Easy to implement refers to ease of deployment to distributed users around the country. The working environments aren’t challenging per se.
14. Please could you clarify what you are looking for where you ask for evidence of conformance to the client’s environment expectations? Does this for example refer to the technical environment on which the Proof of Concept is based, or the business/cultural environment? Does the conformance relate to the Proof of Concept, its outputs or the way in which it was developed? This refers to the technical environment within which the PoC and (ultimately) the end products are delivered. Really we are asking here for compliance with any guidance as and when it is received from our technical architecture colleagues.
15. Could you inform use who the business sponsor within Defra is for this work?
who is the business sponsor for this work, and how does it link to Defra Data and Digital teams?
This project is sponsored by the Defra Data Transformation Programme
16. Are you able to outline the timetable for the process between the application window closing on 31 May and the envisaged start date on 9 July? It would be useful to know when you might expect any written proposals for a second round to be submitted and/or in-person presentations to take place. This is an urgent requirement with the programme moving at pace, so appreciate this is a tight schedule and still provisional: second round requirements to shortlisted suppliers 06 06 18 to be returned 18 06 18 ahead of presentations 21 06 18
17. Could you confirm whether suppliers' first-round responses will be published at any point? We do not publish supplier responses
18. W.r.t the requirement for "Proven capability to deliver data science expertise into projects and share learning". Are you looking for evidence of sharing learnings with the client in the context of the project (e.g. knowledge transfer) or are you looking for key lessons learned from data science project cited? Both apply, but importantly we are looking for the supplier to transfer knowledge into Defra rather than require us to consult with them for everything. In short, part of this is about building Defra’s own *internal* capability in these technologies.
19. It looks like the intention is to provide support to other projects after the first Discovery. What other types of data do you anticipate working with? For example, sensor data, imagery data, natural language, etc? We are not being prescriptive at this stage. Part of what we need to achieve is understanding what data we have and what is useful.
20. Will the proposal response be only for the Discovery Phase? The outcomes/Deliverables at the end of Discovery Phase are not called out. Can you let us know those? We are working on a more detailed specification of what will be required for the proofs of concept, which will be made available to the shortlisted suppliers for the next stage of the tender.
21. The Discovery Phase time line is mentioned as 8 weeks and the budget is mentioned as 1 million. This is a statement not a question, please resubmit if you are seeking clarity on a point.
22. Are the use cases defined under “Who the users are and what they need to do” the only ones to be considered in scope of discovery ? No, these are indicative only. It is expected that user stories will continue to be defined from now and during the proof concept phase.
23. From an AI / ML access to suitable data for training of models will be important. Are these data already available or will preparing these data be part of the work? What about hosting / serving those data? Some activities to obtain the data have started, but this is likely to continue through the initial phases of the PoC. We have assumed that some preparation/cleaning of the data will be required from the vendor. We will be setting up a Defra cloud environment and will provide access to the successful supplier
24. Regarding delivery to front-line users - are you looking for a supplier with the capability to provide ongoing support as well as initial discovery / prototyping / development skills? Note that any support services outside of those needed to deliver this requirement would be the subject of another procurement exercise
25. Are there any specific open source products / frameworks that you are keen to use or explore further? We have not prescribed any tools at this stage and will look forward to working collaboratively with the supplier to select the right tools to deliver the proof of concepts.