Manish Jethwa, chief expertise officer (CTO) at Ordnance Survey (OS), has a career-long ardour for turning geographical information into helpful perception. On the UK’s nationwide mapping service, he’s main the organisation’s improvement of next-generation geospatial applied sciences.
Jethwa studied engineering science on the College of Oxford, the place he targeted on what have been then thought-about area of interest areas, corresponding to robotics, synthetic intelligence (AI) and laptop imaginative and prescient. He subsequently honed this curiosity in a doctorate at MIT, researching how photographs could possibly be used to create fashions of metropolis areas. The chance to work for OS 20 years later felt a bit like coming house.
“A part of that loop had been reconnected,” he says. “I’d been on the earth of making city-scale fashions. After which, earlier than I knew it, I used to be serving to to map the entire nation on the OS.”
Jethwa joined OS in late 2023. He was beforehand CTO at development software program specialist Causeway Applied sciences, a job he’d held since August 2022, following the corporate’s acquisition of asset administration software program agency Yotta. Jethwa had labored at Yotta for 20 years, main product and expertise. As OS CTO, he’s placing his information expertise into follow.
“The explosion in AI means it’s in each organisation’s enterprise technique,” he says. “AI and laptop imaginative and prescient are a core a part of the analysis we do right here, corresponding to robotically extracting options from aerial imagery and at avenue stage.”
For almost all of his 18 months at OS, Jethwa says he’s been extra targeted on the “chief” a part of his CTO function than “expertise” as a result of he’s serving to to drive wide-scale organisational change. That focus is comprehensible for a person who has labored in entrepreneurial companies and joined an organisation that’s 234 years outdated.
“Having come from a product-led enterprise to an organisation that was extra targeted on supply, we’ve needed to change sure issues, and that’s been mirrored in our technique,” he says.
Creating a price chain
Jethwa says his give attention to organisational change has helped break down OS into a price chain of key service pillars that ship advantages to clients.
The organisation has 5 pillars that cowl the next areas: positioning – that means the aptitude across the UK to make correct measurements; information sourcing – aggregating and accumulating info; refinery – including worth to information and merging insights; distribution – pushing information feeds to shoppers; and software – delivering perception to clients.
We’ve constructed up this worth chain from sourcing information to delivering advantages to the client. Clients don’t need lightbulbs, they need mild. On the OS, clients don’t need information, they need insights. Having that mannequin helps outline our strategy Manish Jethwa, Ordnance Survey
“We’ve constructed up this worth chain from sourcing information to delivering advantages to the client,” he says, earlier than utilizing an analogy: “Clients don’t need lightbulbs, they need mild. On the OS, clients don’t need information, they need insights. Having that mannequin helps outline our strategy.”
Jethwa says the result’s a lot of conversations in regards to the worth chain. “I get wheeled out to speak about it,” he says. “But it surely’s a pleasant option to visualise the work and relationships of an organisation of 1,400 individuals. We now have a high-level construction to elucidate how we function.”
Jethwa says he can’t take sole credit score for the organisational change programme. A drive for transformation pre-dated his arrival. His emphasis was on defining clear enterprise companies. Because of this work, from the beginning of April, OS has a service mannequin for operation throughout the 5 key pillars.
“We’ve service leads throughout these areas, and people companies are damaged down into additional parts,” he says. “The great factor about that strategy is that, inside these parts, we are able to handle our expertise stack utilizing the identical framework.”
Specializing in product supply
Jethwa says the brand new organisational construction has enabled a shift from project-based methods of working. Relatively than specializing in smaller initiatives that repair short-term points, the IT workforce seems to fulfill goals as a part of a permanent programme of labor.
“Workers put ahead a programme proposal that can then get accepted,” he says. “You’ve got a finances, you formulate a workforce, you execute in opposition to that. However to ship a long-term imaginative and prescient, you usually have enduring groups that work in your expertise merchandise, so that they perceive the client, they know how you can ship outcomes, they usually additionally handle programs.”
Jethwa says the results of this shift from initiatives to merchandise is a basic change in how OS operates. “These enduring groups have a particular, outlined imaginative and prescient about what they should ship, and the way the outcomes might be handed on to inside or exterior clients,” he says. “They’ve the chance and energy to work out one of the simplest ways of delivering merchandise.”
Jethwa says the groups at all times give attention to a component of transformation. He refers back to the improvement of merchandise inside the sourcing pillar. The professionals take into account how you can gather information and the optimum means of fulfilling that course of, whether or not it’s via drone seize or street-level surveys.
“The strategy to sourcing shifts repeatedly relying on the panorama and the sorts of options we wish to ship,” he says.
“The problem for us is that our information is just not delivered in a single particular 12 months, after which we’re accomplished. As soon as we’ve delivered that information, we now have to offer consistency and ship updates henceforth till we resolve to retire a product, which can take not less than 10 years. It could possibly be longer than that.”
This requirement for steady supply creates a brand new set of challenges, corresponding to coping with the roll-out of rising applied sciences and extracting info and options from supply imagery. Jethwa says AI can assist automate these processes. Nevertheless, care have to be taken to make sure that AI-enabled applied sciences don’t make assumptions about photographs.
“Sure, AI can increase productiveness and effectivity,” he says. “However an automatic technique for analysing photographs could be fooled. So, we should take into account how you can handle uncertainty as we ship insights to clients. That space presents a contemporary problem for us.”
Embracing technological innovation
One other difficulty is technical specs. Jethwa says OS desires to democratise information entry for patrons. Whereas the premise is straightforward, the method should clear some obstacles, notably the truth that most information comes with detailed technical specs.
“The explosion in AI means it’s in each organisation’s enterprise technique. AI and laptop imaginative and prescient are a core a part of the analysis we do right here [at OS], corresponding to robotically extracting options from aerial imagery and at avenue stage”
Manish Jethwa, Ordnance Survey
“A superb instance is perhaps a buyer who says, ‘I wish to know good areas to exit and train’,” he says. “Now, our information, from a technical standpoint, may not reference the phrase train wherever, however it’s going to embrace issues like sports activities stadiums, fields and skateboard parks.”
Jethwa says the standard strategy to constructing a connection between information and technical specs can be to make use of pure language processing. Constructing that mechanism wouldn’t be simple. The excellent news now’s that giant language fashions (LLMs) are excellent for making a relationship between information and specs.
“When you fine-tune a big language mannequin agent, and level it to our information and documentation, you’ll be able to ask a query, corresponding to, ‘Inform me about locations to train’, and have that query instantly translated into an API [application programming interface] request that can say, ‘Listed here are some sports activities services within the buildings information and listed below are some leisure areas within the inexperienced areas information’,” he says.
Jethwa says OS is tapping into the complete breadth of AI applied sciences. The organisation builds lots of its processes on mainstream fashions, endeavor fine-tuning based mostly on inside documentation. Nevertheless, one important difficulty is that many high-profile LLMs are educated on commercially out there information.
To beat this problem, OS faucets into the high-precision UK geography information its groups have collected over many years of labor. On this core space, the place the organisation extracts geographical info, corresponding to roof supplies or biodiversity options, the expertise workforce is constructing foundational fashions from the bottom up.
“The UK panorama may be very totally different from different geographies,” he says. “We all know that coaching fashions on our information will give a richer and extra dependable output than attempting to depend on fashions which were educated on imagery taken instantly from the online.”
Main the way in which
Jethwa says senior executives at OS commonly focus on how individuals would possibly work together with geospatial information throughout the subsequent decade. One factor is already clear – the connection has already modified and can proceed to remodel.
OS has historically been recognized for producing paper-based maps that assist individuals navigate their surroundings. These maps are designed to make it straightforward for individuals to filter the info they require. Jethwa provides the instance of somebody on a hike who can see contour strains on a map and perceive the elevation of their stroll.
When you look forward 5 or 10 years, the connection must be one wherein individuals have way more of a dialog with the map and the interface Manish Jethwa, Ordnance Survey
Nevertheless, the standard relationship is altering. Relatively than specializing in filtering, a big proportion of the organisation’s work is now centred on evaluation. OS is creating AI-enabled processes that clients use to attract insights from information. Jethwa expects the character of this interplay to accentuate.
“When you look forward 5 or 10 years, the connection must be one wherein individuals have way more of a dialog with the map and the interface,” he says. “This relationship might be one the place clients ask a query and get a response again from the map, however they’ll additionally get questions from the map.”
Jethwa says rising expertise is the important thing to unlocking this functionality. Alongside explorations into AI, OS makes use of the Snowflake Market to share open information with public sector organisations and discover new industrial avenues. The precedence now’s guaranteeing OS is able to embrace the following wave of data-led advances.
“We have to guarantee we’ve enabled all of the mechanisms internally. So, how straightforward is it to ask the questions and obtain solutions through APIs? Are we offering the info throughout the fitting sources, whether or not it’s Snowflake, AWS or elsewhere? And are we offering hooks into the fitting shops to make sure that, as fashions are constructed, perception is accessible?” he says.
“We have to prepared the ground as a result of we don’t wish to be late into the sport. If we’re late, there’ll be different suppliers of information that is perhaps much less authoritative, however as a result of they’re simpler to entry, they’ll win out. We must be on the forefront.”
