software/AI specialists

Description

Summary of the work We have developed a ‘data lake’ of structured, machine-readable data and we have built ourselves a suite of visualization tools. Now we are looking for a team of specialists in data analysis, machine learning/AI and software development. Expected Contract Length Approx 2 months Latest start date Wednesday 1 December 2021 Budget Range £40,000-£50,000 Why the Work is Being Done We have developed a ‘data lake’ of structured, machine-readable data and we have built ourselves a suite of visualization tools. You can find out more about our approach to managing our data here (105 mins into the video). https://rssb.videomarketingplatform.co/irsc-2020-webinar-series-learning Now we are looking for a team of specialists in data analysis, machine learning/AI and software development to help us: • determine other analytical applications for our structured data, including pattern recognition and predictive modelling (e.g. machine learning), • design and build a functioning ‘proof-of-concept’ tool, to demonstrate these analytical opportunities to our leadership. Problem to Be Solved The scope of work for the specialists will include the following technical tasks: 1. Advising ORR on potential uses of our structured data and current best practice in these areas, 2. Coding (e.g. writing scripts for data analysis) and software development (e.g. producing basic GUIs), relying exclusively on open-source tools, or those ORR already have access to (Office 365). The key technical challenge is that our data contains c.10,000 different variables, e.g. train performance, number of accidents of a particular type, the weather etc. Each variable has only a limited number of data points (typically 5-40 spatial points; and c.5-10 years of data, most are monthly but some are only annual). This is very different to traditional “big data” datasets which have millions of data points, for a limited number of variables. Our focus is on FLEXIBILITY – allowing our users to look at different variables and solve different problems every day – rather than solving a well defined problem to a high degree of accuracy. Who Are the Users This proof of concept tool will be used by the ORR project team (who are engineering experts with good knowledge of data processing and analysis). They will demonstrate the PoC to senior managers, who are railway experts with limited understanding of data and analysis. Ultimately, applications will be used across our organisation. The majority of users are railway experts with limited understanding of data and analysis. Work Already Done We have already structured our data and built tools for our users to visualize it. We have just started a new project, looking to develop innovative analytical uses for this structured data. This first phase of this project is to build a working ‘proof of concept’, which must be an interactive tool we can use to demonstrate applications to our senior management. Existing Team Very small, very agile, in-house team: currently just 1 project manager and 2 data specialist. Our project deadline is to present a PoC to senior management before the end of this financial year (March 2022). Current Phase Alpha Skills & Experience • Coding, using open-source resources • Developing GUIs, using open-source or Office 365 resources (could be using existing software, e.g. using VBA within Excel or DAX/M within PowerBI - but preferably developing new, bespoke applications) • Rapid prototyping and working with constantly evolving requirements (e.g. AGILE projects, or similar) Nice to Haves • Developing GUIs as new software applications, from scratch. • Previous work relating to UK transport or other public sector. Work Location Either: remotely, communicating over MS Teams, or Canary Wharf, London. Working Arrangments The specialists will provide contingent labour, working directly for the project team. Full remote working, full in-office working or a hybrid - to be proposed by the applicant, Daily communication with the project team, where the Project Manager will assign tasks and priorities. This role will be time-limited (e.g. 40 working days – to be agreed before starting, based on the specialists’ rates and availability), rather than delivering a specified output. The outputs of the project will be agreed between the specialist and the project manager as the project progresses, based on the technical options presented by the specialist. No. of Suppliers to Evaluate 5 Proposal Criteria • Show examples of innovative analytical solutions, including pattern recognition and machine learning – with a focus on providing flexible interfaces for users (i.e. GUIs, rather than outputting “42”). • Knowledge of global best practice – and ideas to push beyond this. Cultural Fit Criteria • The individuals must be open-minded, innovative, and able to take on new ideas and change course very quickly. • Can work with users with low technical expertise. Payment Approach Capped time and materials Assessment Method Presentation Evaluation Weighting Technical competence 60% Cultural fit 20% Price 20% Questions from Suppliers 1. Please advise on the application process i.e. are proposals due following the initial questionnaire submission (if successful)? Please confirm rates are not to be submitted anywhere at this initial questionnaire stage Yes written proposals and possibly presentation are only due after shortlisting along with price. 2. Was there was a discovery phase and if yes who has delivered it? Discovery (i.e. defining target users and their requirements) has already been done by our in-house project team. 3. Are there any relevant material we could look at? We included a weblink in the DOS ad, which includes a short presentation on our approach and an early demo of our visualisation tools. 4. Is the provided budget inclusive or exclusive of VAT? Exclusive 5. Will you consider the provision of offshore expertise in a submission. If by this question you mean can Non-UK based suppliers apply for this opportunity. Yes as they are permitted to apply to join the DOS framework and as such can bid for any published opportunity where they consider they have the capability to deliver the requirements. 6. In the existing team, 2 data specialists are referenced. What is their existing technical skillset? Particularly in regard to use of open-source tooling. In-house project resources include a project manager (with a PhD in 3D-image analysis and good coding skills in Visual Basic and MATLAB, plus good skills in PowerBI), and access to resources from our in-house “data science” and “business intelligence” teams (with a range of skills across VBA, Python, R and PowerBI). The “gaps” in our skillset, which we require from the applicant, are experience of, and the necessary coding skills to do: pattern recognition, machine learning, developing bespoke GUIs (beyond what’s possible in PowerBI etc) and insight into the frontiers of “clever stuff you can do with data”. 7. Is there an incumbent supplier that ORR is working with for this project? No 8. Who is reviewing and marking the responses? The ORR project manager & engineering colleagues to provide a balanced and diverse assessment. 9. Do you have a measure of the accuracy of the NLP (natural language processing) systems for pulling information from different sources? To date we have only used NLP for simple tasks, on relatively well structured documents (e.g. extracting paragraphs from templated incident logs, then identifying the location of the incident by looking for phrases like “at [recognisable location name]” or “between [] and []”. We measured the accuracy of this based on visually inspecting a sample of X%count of “how many failed to return anything recognisable” (which we can mitigate with approx. results, e.g. “it didn’t find an exact location for the incident, but we can tag it to the Wales region, based on the first line of the text 10. The RFP indicates that a data lake and visualisation tools already exist – what technology stack/platform has this been created on for example Azure or AWS? Data is stored in Azure Blob and our visualisation tool uses PowerBI. 11. How do you account for potentially bad accuracy in the data being extracted? Given the range of data sources, there will be inaccurate data. We have a separate team who quality assure incoming data and maximise accuracy before it gets anywhere near this project. This project is not responsible for improving the accuracy of the input data, nor fine tuning the tools to recognise inaccuracies in a particular dataset – this project is all about developing flexible tools, into which we can plug any data we want (provided its in a consistent format), then provide a user interface which is easy-to-interpret, so users can make their own judgement on the accuracy of inputs/outputs. 12. Can you please describe a few examples of the data currently held within your data lake? - 4-weekly data on percentage of trains on time (within 5 mins, within 10 mins etc) for every train operating company, going back decades- 4-weekly data on causes of delays (e.g. trespassers, flooding, power failures), for every local maintenance area (York, Derby etc…39 of them), for 5 years - Monthly rainfall data for different areas, going back decades- 10,000 variables like the examples above, covering performance, financials, asset failures, safety, environmental, weather etc - We also have some text-based data, including c.45,000 descriptions of individual incidents, going back c.5 years 13. Can you provide information regarding subcontracting. Any supplier on the framework (in their capacity as prime supplier) may enter into subcontracting arrangements for the delivery of part of the requirement. This supplier should detail this in their written proposal along with the percentage of delivery allocated to the subcontractor.The supplier will only subcontract with the prior written approval of the buyer.The subcontractors can be non-UK based. 14. Could you please define the meaning of ‘from scratch’ in the context of Developing GUIs as new software applications? For example, the end product might be an .exe file, or a web application – when the user opens it, it presents a GUI asking the user for input parameters, then it processes the necessary data, then it presents the user with a GUI. This is as opposed to the whole process being done inside an existing piece of software, e.g. PowerBI, Excel etc. 15. You are open to on-site, hybrid or remote teams. However, do you have a preference? The project manager’s preference would be remote (with daily video calls) and occasional meetings in our London office (e.g. once a week), as this can help with brainstorming on whiteboards, grabbing other colleagues to ask questions etc. But we are open and we would like applicants to suggest the system they think will produce the best results. 16. Are you happy to provide the Discovery report to shortlisted bidders, to help us tailor our proposed approach to the challenge in a manner that adds most value? Not as a formal report, but this is a good idea and we are willing to provide a short summary (e.g. a few powerpoint slides) to shortlisted bidders, so they can use it to prepare for the interview/presentation stage.