\n
![\"\"](\"https:\/\/www.europesays.com\/lv\/wp-content\/uploads\/2026\/01\/1BRL2343-Large.jpeg\")
<\/a>\u201cDatakom\u201d biznesa att\u012bst\u012bbas vad\u012bt\u0101js Edijs Tanons\n<\/p>\nK\u0101ds \u201cDatakom\u201d bija pag\u0101ju\u0161ais gads? K\u0101das interesantas tendences nov\u0113roj\u0101t?<\/p>\n
Pag\u0101ju\u0161ais gads bija izaicino\u0161s. IT tirg\u016b notika daudz un straujas p\u0101rmai\u0146as. Neskatoties uz to, \u201cDatakom\u201d finan\u0161u rezult\u0101ti ir labi. Esam \u013coti daudz ener\u0123ijas un resursu ieguld\u012bju\u0161i jaunos biznesa virzienos. P\u0113rn savu darb\u012bbu akt\u012bvi izv\u0113rs\u0101m Baltij\u0101. Piem\u0113ram, Lietuv\u0101 s\u0101k\u0101m pied\u0101v\u0101t pakalpojumu \u201cTiki-Taka PAY\u201d. T\u0101pat turpin\u0101j\u0101m att\u012bst\u012bt eso\u0161os produktus. Viens no tiem ir IT \u0101rpakalpojums \u201cWorry-Free IT\u201d, kas ir visaptvero\u0161s IT uztur\u0113\u0161anas pakalpojums. B\u016btiski, ka p\u0113rn ar\u012b izveidoj\u0101m partner\u012bbu ar \u201cNvidia\u201d. T\u0101s ietvaros uz\u0146\u0113mumiem pied\u0101v\u0101jam \u201cNvidia\u201d superdatorus. Pirm\u0101s iek\u0101rtas pie klientiem non\u0101ca jau p\u0113rn\u0101 gada beig\u0101s.\n<\/p>\n
Vai no klientu puses j\u016btat vajadz\u012bbu p\u0113c produktiem, kuru pamat\u0101 ir m\u0101ksl\u012bgais intelekts?<\/p>\n
Manupr\u0101t, tas m\u0101ksl\u012bg\u0101 intelekta burbulis ir milz\u012bgs. Vienlaikus MI sp\u0113jas pag\u0101ju\u0161\u0101 gada laik\u0101 ir b\u016btiski att\u012bst\u012bju\u0161\u0101s. Ja 2025. gada s\u0101kum\u0101 MI v\u0113l bija sal\u012bdzino\u0161i neprofesion\u0101ls un sniedza aplamas atbildes, tad gada beig\u0101s precizit\u0101te un kvalit\u0101te jau bija krietni augst\u0101ka. Kopum\u0101 \u0161obr\u012bd MI paveic uzdevumus lab\u0101k nek\u0101 juniora l\u012bme\u0146a darbinieks.\n<\/p>\n
Interese no m\u016bsu klientu puses par MI risin\u0101jumiem ir. Vienlaikus nov\u0113rojam tipisk\u0101 latvie\u0161a piesardz\u012bbu un v\u0113lmi pa maziem sol\u012b\u0161iem ieiet \u0161aj\u0101 jom\u0101 \u2013 lai k\u0101ds cits pam\u0113\u0123ina, tad m\u0113s.\n<\/p>\n
V\u0113l p\u0113rn\u0101 gada vid\u016b t\u0101di superdatori, k\u0101dus \u201cDatakom\u201d pied\u0101v\u0101 saviem klientiem tagad, nebija tik pieejami un maks\u0101ja simtiem t\u016bksto\u0161u. Uz\u0146\u0113mumiem t\u0101 \u0161\u0137ita p\u0101r\u0101k liela invest\u012bcija. Tr\u016bka ar\u012b izpratnes, k\u0101p\u0113c t\u0101das ier\u012bces ir vajadz\u012bgas. Liel\u0101ka interese no klientu puses par\u0101d\u012bj\u0101s p\u0113c tam, kad \u201cDatakom\u201d maij\u0101 pazi\u0146oja, ka veido in\u017eenieru komandu un uz saviem datiem ir apm\u0101c\u012bjis MI mode\u013cus. \u0160odien jau varam konkr\u0113ti par\u0101d\u012bt, k\u0101 superdators str\u0101d\u0101 praks\u0113.\n<\/p>\n
Uz\u0146\u0113mumiem ir savi dati, kurus negribas vai nedr\u012bkst s\u016bt\u012bt uz \u0101rzem\u0113m, t\u0101p\u0113c pieaug vajadz\u012bba p\u0113c pa\u0161u uztur\u0113tiem serveriem un risin\u0101jumiem. Vai tas bija viens no iemesliem, k\u0101p\u0113c izveidoj\u0101t sadarb\u012bbu ar \u201cNvidia\u201d?\n<\/p>\n
Tie\u0161i t\u0101. Tu jau b\u016bt\u012bb\u0101 atbild\u0113ji uz \u0161o jaut\u0101jumu. Liel\u0101s korpor\u0101cijas parasti sola: \u201cM\u0113s visu izdar\u012bsim j\u016bsu viet\u0101\u201d. Praks\u0113, kad izlasi l\u012bgumu, k\u013c\u016bst skaidrs, ka, ja kaut kas pazud\u012bs, atg\u016bt savus datus b\u016bs gandr\u012bz neiesp\u0113jami. Tu esi mazs sp\u0113l\u0113t\u0101js sal\u012bdzin\u0101jum\u0101 ar \u0161\u012bm lielaj\u0101m korpor\u0101cij\u0101m. Paties\u012bb\u0101 tu vi\u0146iem nemaz neinteres\u0113. Tie\u0161i t\u0101p\u0113c lok\u0101las sist\u0113mas k\u013c\u016bst arvien piepras\u012bt\u0101kas. Ne visiem t\u0101s ir vajadz\u012bgas, bet ir nozares, kur tas ir \u013coti aktu\u0101li.\n<\/p>\n
K\u0101dus \u201cNvidia\u201d superdatorus pied\u0101v\u0101jat klientiem?<\/p>\n
Analiz\u0113jot klienta v\u0113lmes un vajadz\u012bbas, piemekl\u0113jam atbilsto\u0161\u0101ko iek\u0101rtu. Pats maz\u0101kais superdators, ko pied\u0101v\u0101jam, ir \u201cNvidia DGX Spark\u201d. Tas maks\u0101 ap 5000 eiro un darbojas ar \u201cNvidia\u201d programmat\u016bru, kas \u013cauj uz\u0146\u0113mumiem apm\u0101c\u012bt MI mode\u013cus, izmantojot iek\u0161\u0113jos datus.\n<\/p>\n
\u201cNvidia\u201d ir pla\u0161s kl\u0101sts ar superdatoriem \u2013 gan maz\u0101ki un liel\u0101ki, gan l\u0113t\u0101ki un l\u012bdz miljons dol\u0101riem v\u0113rti. \u201cNvidia DGX Spark\u201d ir nedaudz liel\u0101ks par \u201cApple Mac Mini\u201d. Tam ir 2 petaflopu jauda jeb tas ik sekundi var veikt 2 kvadriljonus skait\u013co\u0161anas darb\u012bbu. Tehnolo\u0123iju izst\u0101d\u0113 \u201cCES 2026\u201d \u201cNvidia\u201d prezent\u0113ja superdatoru, kuram ir divarpus rei\u017eu liel\u0101ka jauda nek\u0101 \u201cDGX Spark\u201d. Kad zini, k\u0101 darbojas \u0161ie mazie superdatori, saproti, ko tas noz\u012bm\u0113.\n<\/p>\n
K\u0101dus MI mode\u013cus var darbin\u0101t ar \u201cNvidia DGX Spark\u201d?<\/p>\n
\u201cNvidia DGX Spark\u201d sp\u0113j darbin\u0101t modeli l\u012bdz 400 miljardiem parametru, kas ir apm\u0113ram \u201cOpenAI GPT-4\u201d izm\u0113ra modelis. Jaun\u0101kajiem publiski pieejamajiem MI mode\u013ciem ir pat triljons parametru. Lai t\u0101du darbin\u0101tu, nepiecie\u0161ams liel\u0101ks superdators.\n<\/p>\n
Uz\u0146\u0113mumiem liel\u0101koties vajag specifiskus MI mode\u013cus, kas dom\u0101ti konkr\u0113tu uzdevumu veik\u0161anai. Uz \u201cDGX Spark\u201d \u0161os specifiskos mode\u013cus var salikt t\u0101d\u0101 k\u0101 \u0137\u0113d\u012bt\u0113. Piem\u0113ram, ir galvenais MI modelis, kas \u013coti labi run\u0101 latviski, un papildus tam vair\u0101ki maz\u0101ki mode\u013ci, kas veic att\u0113lu, video anal\u012bzi utt. Visi \u0161ie mode\u013ci str\u0101d\u0101 k\u0101 viena komanda. Ar galveno modeli vari sarun\u0101ties \u010data form\u0101, uzdot jaut\u0101jumus un sa\u0146emt atbildes. Piem\u0113ram, iedod modelim analiz\u0113t 10 stundu garu nov\u0113ro\u0161anas video un jau p\u0113c min\u016btes sa\u0146em atbildes uz jaut\u0101jumiem.\n<\/p>\n
Cik viegli ir izveidot kl\u0101steri no \u201cDGX Spark\u201d vai cit\u0101m \u201cNvidia\u201d iek\u0101rt\u0101m?<\/p>\n Vienu iedarbin\u0101t ir vienk\u0101r\u0161i. Divus superdatorus salikt kop\u0101 ir jau diezgan sare\u017e\u0123\u012bti. Ir \u013coti j\u0101iedzi\u013cin\u0101s instrukcij\u0101, lai saprastu, k\u0101 tos pareizi sakonfigur\u0113t, lai tie darbotos k\u0101 viens vesels. Salikt kop\u0101 divus \u201cNvidia DGX Spark\u201d ir lo\u0123isks solis, jo uz tiem var palaist gana lielus MI mode\u013cus,\u00a0 ar\u012b \u0101trums ir \u013coti labs.\n<\/p>\n \u201cNvidia\u201d saka, ka iesp\u0113jams sasl\u0113gt ne vair\u0101k k\u0101 divus \u0161\u0101dus superdatorus. Internet\u0101 gan ir atrodama inform\u0101cija, ka var izveidot kl\u0101steri pat no \u010detriem \u201cDGX Spark\u201d. Ta\u010du tas neb\u016bs finansi\u0101li pr\u0101t\u012bgi. Ja nepiecie\u0161ama liel\u0101ka jauda, lab\u0101k uzreiz pa\u0146emt jaud\u012bg\u0101ku superdatoru, kas maks\u0101s tikpat vai mazliet l\u0113t\u0101k.\n<\/p>\n K\u0101da ir \u201cDatakom\u201d pievienot\u0101 v\u0113rt\u012bba?<\/p>\n \u201cDatakom\u201d ir \u201cNvidia\u201d ofici\u0101lais partneris. Tas noz\u012bm\u0113, ka mums ir pieredze un zin\u0101\u0161anas tirgot un piemekl\u0113t klientam atbilsto\u0161\u0101kos \u201cNvidia\u201d produktus. \u0160\u012b partner\u012bba \u201cDatakom\u201d ar\u012b dod piek\u013cuvi \u201cNvidia\u201d programmat\u016brai, dzel\u017eiem, in\u017eenieriem un izp\u0113tes centriem, ar kuriem mums ir tie\u0161s kontakts. Ja kaut ko nesaprotam, pajaut\u0101jam vi\u0146iem. Manupr\u0101t, milz\u012bga pievienot\u0101 v\u0113rt\u012bba no \u0161\u012bs sadarb\u012bbas ir tie\u0161i zin\u0101\u0161anas, ko uzreiz ieg\u016bstam no \u201cNvidia\u201d. Tas \u013cauj \u201cDatakom\u201d pal\u012bdz\u0113t klientam saprast, kur\u0101 virzien\u0101 skat\u012bties, k\u0101 uzb\u016bv\u0113t risin\u0101jumu t\u0101, lai dzelzis un programmat\u016bra sniegtu liel\u0101ko pievienoto v\u0113rt\u012bbu biznesam.\n<\/p>\n Cik liela probl\u0113ma \u0161obr\u012bd ir dzel\u017eu pieg\u0101de?<\/p>\n Ir \u013coti j\u016btams ier\u012b\u010du defic\u012bts. Decembr\u012b sa\u0146\u0113m\u0101m vair\u0101kus \u201cNvidia DGX Spark\u201d mazos superdatorus. Ta\u010du tiem klientiem, kuri v\u0113las specifisk\u0101ku risin\u0101jumu, \u0161obr\u012bd pieg\u0101de j\u0101gaida tr\u012bs l\u012bdz se\u0161us m\u0113ne\u0161us. Pie\u0146emu, ka Amerikas lielie datu centri ir izpirku\u0161i tirgu sausu. T\u0101s r\u016bpn\u012bcas, kur\u0101s ra\u017eo superdatorus, ar\u012b neaug glu\u017ei k\u0101 s\u0113nes p\u0113c lietus.\n<\/p>\n K\u0101di priek\u0161darbi klientam ir j\u0101veic, lai \u201cNvidia DGX Spark\u201d var\u0113tu pilnv\u0113rt\u012bgi izmantot?<\/p>\n T\u012bkls ir vien\u012bgais, kas var tehniski ierobe\u017eot \u0161\u012b superdatora pilnv\u0113rt\u012bgu izmanto\u0161anu. Klientam ir nepiecie\u0161ams jaud\u012bgs t\u012bkls. Piem\u0113ram, \u201cNvidia DGX Spark\u201d ir vair\u0101ki t\u012bkla piesl\u0113gumi \u2013 jaud\u012bg\u0101kais atbalsta 200 gigabitus, bet otrs ir ar 10 gigabitu \u0101trumu.\n<\/p>\n Pirms ieg\u0101d\u0101ties \u0161\u0101du iek\u0101rtu klientam ir b\u016btiski apzin\u0101ties m\u0113r\u0137i, k\u0101dus uzdevumus ar to v\u0113las veikt. Uz \u201cDGX Spark\u201d nav v\u0113rts darbin\u0101t \u201cExcel\u201d vai \u201cPowerPoint\u201d. \u0160\u012b iek\u0101rta ir dom\u0101ta skait\u013co\u0161anas oper\u0101cij\u0101m, ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013cu apm\u0101c\u012bbai, lielo datu apstr\u0101dei re\u0101llaik\u0101.\n<\/p>\n K\u0101d\u0101s jom\u0101s var izmantot \u0161\u0101dus superdatorus?<\/p>\n Noteikti datu anal\u012btik\u0101 da\u017e\u0101d\u0101s nozar\u0113s. Pirmie \u201cDatakom\u201d klienti, kas nopirka superdatorus, bija tie, kas p\u0113ta un mekl\u0113 kr\u0101pniec\u012bbas, korupcijas riskus. Vi\u0146iem bija skaidra v\u012bzija par \u0161o ier\u012b\u010du pielietojumu, k\u0101 ar\u012b \u013coti liels analiz\u0113jamo datu apjoms. Finan\u0161u risku jom\u0101 superdators var analiz\u0113t datus ne tikai izlases sec\u012bb\u0101, bet piln\u012bgi visu uzreiz un re\u0101llaik\u0101.\n<\/p>\n Superdatoriem ir labs pielietojums ar\u012b publiskaj\u0101 sektor\u0101 darb\u0101 ar liela apjoma dokumentiem un datiem. \u013boti v\u0113rt\u012bga pieredze \u201cDatakom\u201d bija ar lielu valsts iest\u0101di, kurai ir milz\u012bga dokumentu aprite. Pilotprojekta ietvaros par\u0101d\u012bj\u0101m, k\u0101 ar MI pal\u012bdz\u012bbu to visu apstr\u0101d\u0101t, izanaliz\u0113t, izveidot kopsavilkumus. T\u0101 bija \u013coti veiksm\u012bga sadarb\u012bba, ko apstiprin\u0101ja ar\u012b klients, kas \u0161ogad nopietni s\u0101ks izmantot superdatoru ikdienas darb\u0101.\n<\/p>\n Ar\u012b universit\u0101tes, p\u0113tniec\u012bbas instit\u016bcijas izr\u0101da interesi par superdatoriem, ko v\u0113las izmantot p\u0113tniec\u012bb\u0101, model\u0113\u0161an\u0101. \u0160\u012bs iek\u0101rtas \u013cauj daudz \u0101tr\u0101k izstr\u0101d\u0101t ar\u012b jaunus sist\u0113mu interfeisus. Agr\u0101k bija vajadz\u012bgs sist\u0113mu anal\u012bti\u0137is, kur\u0161 p\u0113t\u012bja biznesa pras\u012bbas, ierobe\u017eojumus un dom\u0101ja, k\u0101 veidot atbilsto\u0161u interfeisu un funkcijas. Tagad ar MI pal\u012bdz\u012bbu to visu var izdar\u012bt p\u0101ris stundu laik\u0101. Vienlaikus joproj\u0101m ir vajadz\u012bgs zino\u0161s cilv\u0113ks, kur\u0161 p\u0101rbauda, vai viss izdar\u012bts pareizi un nav nepiecie\u0161ami labojumi.\n<\/p>\n Kur klientiem v\u0113rsties, ja\u00a0 ir vajadz\u012bba p\u0113c \u0161\u0101da superdatora?<\/p>\n \u201cDatakom\u201d m\u0101ksl\u012bg\u0101 intelekta jomu redz k\u0101 virzienu, kur\u0101 j\u0101dodas ne tikai uz\u0146\u0113mumiem Latvij\u0101, bet biznesiem vis\u0101 pasaul\u0113. M\u016bsu m\u0113r\u0137is \u0161ogad ir papla\u0161in\u0101t savu \u201cAI Datacenter Engineers\u201d in\u017eenieru komandu un pied\u0101v\u0101t \u0161os pa\u0161us pakalpojumus Baltij\u0101 un Eirop\u0101. Uz\u0146\u0113mumi Latvij\u0101 m\u016bs var atrast m\u0101jaslap\u0101 datakom.lv<\/a>. \u0100rvalst\u012bs darbojamies ar z\u012bmolu \u201cAI Datacenter Engineers\u201d un sazin\u0101ties var caur m\u0101jaslapu aidatacenterengineers.com<\/a>. Ja ir ideja, bet nav l\u012bdz galam izdom\u0101ts, k\u0101 to realiz\u0113t, uzrakstiet. Satiksimies un saprat\u012bsim, vai t\u0101 var j\u016bsu biznesam nest v\u0113rt\u012bbu.<\/p>\n","protected":false},"excerpt":{"rendered":"Skait\u013co\u0161ana, datu anal\u012btika, m\u0101ksl\u012bg\u0101 intelekta mode\u013cu apm\u0101c\u012b\u0161ana un darbin\u0101\u0161ana, da\u017e\u0101du prototipu veido\u0161ana \u2013 tie ir tikai da\u017ei no…\n","protected":false},"author":2,"featured_media":102932,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[77,76,35,39,38,36,37,34,40],"class_list":{"0":"post-103062","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-bizness","8":"tag-bizness","9":"tag-business","10":"tag-latvia","11":"tag-latvian","12":"tag-latviesu","13":"tag-latviesu-valoda","14":"tag-latviesuvaloda","15":"tag-latvija","16":"tag-lv"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@lv\/115887978743965887","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/posts\/103062","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/comments?post=103062"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/posts\/103062\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/media\/102932"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/media?parent=103062"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/categories?post=103062"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/lv\/wp-json\/wp\/v2\/tags?post=103062"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}<\/a>\n<\/p>\n