Algorithm for the masses

Daniel Hulme is the founder and chief exec­u­tive of a young British com­pa­ny spe­cial­is­ing in opti­mi­sa­tion algo­rithms and ser­vices. It’s akin to a maths think-tank. Satalia’s staff of aca­d­e­mics and arti­fi­cial intel­li­gence or AI experts har­ness a library of algo­rithms to help com­pa­nies solve the world’s most dif­fi­cult prob­lems. It was recog­nised this year on the elite Gart­ner Cool Ven­dors in Data Sci­ence list, the only UK com­pa­ny cho­sen.

It helped a major UK tel­co embark on a big upgrade pro­gramme. Cop­per lines would be replaced with fibre. With thou­sands of miles of trunk­ing to con­sid­er, the tel­co need­ed to cal­cu­late the best pos­si­ble routes. Even tiny improve­ments to plan the fibre routes would mean huge cash sav­ings. The prob­lem? The num­ber of pos­si­ble routes to con­sid­er quick­ly spi­rals out of con­trol. There are tril­lions upon tril­lions of con­ceiv­able sequences. It’s insane­ly hard to com­pute.

In fact, the prob­lem will be famil­iar to any­one with an engi­neer­ing back­ground. It’s a twist on the trav­el­ling sales­man prob­lem taught to every com­put­er sci­ence and advanced maths stu­dent.

Satalia ran the tel­co chal­lenge through its algo­rithms library they call the Solve Engine, which uses machine-learn­ing to select the best algo­rithms to solve opti­mi­sa­tion prob­lems. The Solve Engine found a solu­tion 1,000 times faster and 20 per cent bet­ter than the exist­ing approach.

We can tack­le any type of opti­mi­sa­tion prob­lem, which usu­al­ly involves some com­bi­na­tion of plan­ning, sched­ul­ing and rout­ing

The exam­ple is far from a one-off. Satalia is at the fore­front of bring­ing algo­rithms to the mass­es. It has a grow­ing port­fo­lio of suc­cess sto­ries. In each case Satalia uses algo­rithms to solve an opti­mi­sa­tion prob­lem too com­plex for the client to han­dle.

“We can tack­le any type of opti­mi­sa­tion prob­lem, which usu­al­ly involves some com­bi­na­tion of plan­ning, sched­ul­ing and rout­ing,” says Satalia founder Daniel Hulme. “We have a library of thou­sands of algo­rithms and deep exper­tise in opti­mi­sa­tion, which means we can help clients with almost any com­plex task. We deal with a wide vari­ety of requests, includ­ing work­force sched­ul­ing, plan­ning mobile phone masts, iden­ti­fy­ing bugs in soft­ware and cre­at­ing the best vehi­cle deliv­ery routes.”

Even small prob­lems can be blitzed by the Satalia team. A uni­ver­si­ty asked Satalia to cre­ate a timetable to fit stu­dent pref­er­ences and lec­tur­er avail­abil­i­ty. The process cre­at­ed the most effi­cient timetable pos­si­ble, while cut­ting the task time by 60 per cent.

Dr Hulme is evan­gel­i­cal about the poten­tial for com­pa­nies of all sizes to embrace algo­rithms to solve every­day prob­lems. Back in 2007 he devel­oped the idea of opti­mi­sa­tion-as-a-ser­vice while doing his doc­tor­al the­sis in AI at Uni­ver­si­ty Col­lege Lon­don. He has since become a glob­al fig­ure­head of the algo­rithm rev­o­lu­tion, speak­ing at TEDx talks and indus­try events on how algo­rithms can be har­nessed.

His mis­sion is to bring algo­rithms to the mass­es and help busi­ness lead­ers realise how crit­i­cal opti­mi­sa­tion is to the suc­cess of their com­pa­ny. “We’ve all heard of big data and advanced ana­lyt­ics,” he says. “The basic lev­el is descrip­tive ana­lyt­ics, which is just the visu­al­i­sa­tion of data. Lots of firms do that. Then there’s pre­dic­tive ana­lyt­ics, which is about mak­ing fore­casts. A small num­ber do that.

“We are offer­ing pre­scrip­tive ana­lyt­ics. This is a new field in which algo­rithms rec­om­mend spe­cif­ic actions. With pre­scrip­tive ana­lyt­ics you now know what to do with the insights extract­ed from data.”

One of Satalia’s biggest suc­cess­es is the opti­mi­sa­tion of the deliv­ery routes for a lead­ing inter­na­tion­al retail­er. The client under­stood the poten­tial cost-sav­ings of using algo­rithms to com­pute the best jour­neys for deliv­ery vehi­cles.

Dr Hulme says: “Sup­pose you have to deliv­er pack­ages around in a city. If you have 24 pack­ages to deliv­er there would be 24x23x22x21 and so on (620,448,401,733,239,439,360,000) pos­si­ble routes. Even if a super­com­put­er could cal­cu­late a mil­lion routes every sec­ond it would still take 20 bil­lion years – longer than the age of the uni­verse – to deter­mine the best one. Add anoth­er deliv­ery and it’s 25x20 bil­lion years (500 bil­lion years). This is why algo­rithms are so impor­tant in the rapid­ly grow­ing com­plex world of busi­ness.”

For large cor­po­rates, we build inte­grat­ed end-to-end AI solu­tions; for small com­pa­nies, we take away the need to have expen­sive data-sci­ence and opti­mi­sa­tion teams

So can any com­pa­ny approach Satalia with a prob­lem to solve? “Absolute­ly,” says Dr Hulme. “We’ll take a look at any­thing and we’re moti­vat­ed to solve the biggest prob­lems. For large cor­po­rates, we build inte­grat­ed end-to-end AI solu­tions; for small com­pa­nies, we take away the need to have expen­sive data-sci­ence and opti­mi­sa­tion teams.”

Satalia has clients from the Unit­ed States to Japan, in mul­ti­ple sec­tors. “We’ll soon be launch­ing a pub­lic ver­sion of the Solve Engine,” Dr Hulme reveals. “Cus­tomers will buy solve time by the hour and every time they sub­mit a prob­lem, it will use up some time, sim­i­lar to mobile phone pay-as-you-go cred­its.” The pric­ing mod­el is flex­i­ble. Jobs can also be priced per prob­lem, through an enter­prise-lev­el con­tract or with a per­ma­nent on-site pres­ence. A freemi­um mod­el is also avail­able.

Aware­ness of algo­rithms is soar­ing. A lot of com­pa­nies look at exam­ples, such as Amazon’s prod­uct rec­om­men­da­tion ser­vice or an airline’s dynam­ic pric­ing engine, and won­der whether it is plau­si­ble to cre­ate some­thing sim­i­lar in their own enter­prise. Satalia makes it pos­si­ble.

“Too many com­pa­nies are either not using algo­rithms or using vanil­la in-house ver­sions that real­ly aren’t fit for pur­pose,” says Dr Hulme. “Our Solve Engine and expert ser­vices means they can com­plete­ly future-proof their opti­mi­sa­tion.”

Satalia’s algo­rithm library is reg­u­lar­ly updat­ed by the aca­d­e­m­ic com­mu­ni­ty, using the lat­est inno­va­tions in opti­mi­sa­tion. This allows com­pa­nies to make the best deci­sions from data and focus on their core busi­ness.

Dr Hulme is reg­u­lar­ly invit­ed to events and con­fer­ences to speak about the role of arti­fi­cial intel­li­gence in tomorrow’s world. “As a com­pa­ny we don’t do cold call­ing,” he says. “We raise aware­ness about the huge impact of opti­mi­sa­tion and algo­rithms, and how to solve incred­i­bly hard prob­lems. Then when com­pa­nies realise what we can do for them, we become part­ners.”

To find out more vis­it Satalia.com or e‑mail Daniel Hulme at daniel@satalia.com

Top image — Dr Daniel Hulme, chief exec­u­tive: “I love talk­ing about what is, and what is not, arti­fi­cial intel­li­gence”