Kuchungechunge lomlobi we-Spotlight, abahleli be-TDS baxoxa namalungu omphakathi wethu ngendlela yabo yomsebenzi ku-Data Science ne-AI, ukubhala kwabo, nemithombo yabo yokuphefumlelwa. Namuhla, siyajabula ukuhlanganyela izingxoxo zethu nazo Vyacheslav efimov.
I-VyachesLav ngunjiniyela oPhakathi oPhakathi Wokufunda Okugxile ku-NLP nombono wekhompyutha. Enye yezifiso zakhe yenza okuqukethwe kwezemfundo: I-Vyacheslav ishicilele izindatshana ezingaphezu kuka-60 ibheke kwisayensi yedatha, ichaza imiqondo eyinkimbinkimbi ngamagama alula, ukuhlinzeka ngezinto eziqondayo.
Ubhale izindatshana eziningi futhi ezichazayo ku-TDS. Ingabe ukufundisa izisekelo kuguqulwe ukuthi uklama noma ulungisa amasistimu wangempela we-debug?
Ngiyabona ukuxhumeka phakathi kokuningi ngifundisa okuthile, kungcono ngiyakuqonda. Empilweni yangempela, lapho ngibhala i-athikili entsha, ngilwela ukungena ngemininingwane emincane ngenkathi ngigcina incazelo ilula kubafundi bami. Ukugcizelela imininingwane ngale ndlela kungizwisisela kangcono ukuhamba komsebenzi kwama-algorithms.
Ngaleyo ndlela, noma nini lapho ngihlangana nephutha kwelinye lama-algorithms asetshenziswa emsebenzini engibhale ngalo i-athikili phambilini, kunethuba eliphakeme lokuthi ngizothola isixazululo senkinga yedwa. Ukusuka kolunye umbono, lapho ngibhala i-athikili ngesihloko esingajwayelekile futhi ngiyihlole, kukhulisa ukuzethemba kwami lapho ngifaka leyo nkinga yalo, njengoba sengivele ngazi isilinganiso saso sokwenza isicelo, izinzuzo, ubunzima, nemininingwane ethile.
Ngale ndlela, ngingaqhamuka nezisombululo zangempela ezingabonakali kwabanye futhi zibuyise ukukhetha kwami kwabanye abasebenza nabo, abaphathi, noma ababambiqhaza. Lolo lwazi luyigugu kimi.
Ngamamodeli amaningi kangaka avela nsuku zonke, kulula ukuzizwa ucwayiswe ngokuphelele. Unquma kanjani ukuthi kufanelekile 'i-dive ejulile' nokuthi yini oyitholayo nje '? Ngabe isu lakho lokuphatha lokhu kuguqulwe muva nje?
Namuhla, empeleni sinenqwaba yamamodeli namathuluzi avela nsuku zonke. Kulula ukuzizwa ulahlekile lapho ungaqiniseki ngokuthi yini okufanele uyiphishekele ngokulandelayo.
Ngesikhathi esinqunyelwe, ngivame ukuhlanganisa ukujula ezihlokweni ezingasebenza emsebenzini noma kumaphrojekthi ami. Lokhu kunginikeza ukuzethemba okwengeziwe lapho kufanele ngiveze noma ngichaze imiphumela yami.
Amabhizinisi ajwayele ukufuna ukufeza imiphumela yokusebenza ngokushesha okukhulu. Lokhu kungenye yezizathu zokuthi kungani, ezihlokweni zami, ngigxila kakhulu emibonweni yethiyori, njengoba angikwazi ukunikela ngesikhathi sami emsebenzini ukuya ekujuleni kwethiyori.
Ngale ndlela, nginenhlanganisela ephumelelayo yesipiliyoni esisebenzayo emsebenzini kanye nokuqonda kwethiyori kubhulogi lami. Zombili lezi zinto zibalulekile kusosayensi wedatha onamakhono.
Uqhubekele e-AI Hackathons. Yini oyifundile ekutholeni imisele enjalo eqinile? Ngabe kukuphoqa ukuthi ube ngcono emapholokhini e-Scoping noma uthathe isinqumo ngemodeli? Futhi uyazithola usebenzisa noma yiziphi zalezi 'zifundo ze-hackathon' lapho ukhipha umbono omusha kusuka ekuqaleni?
AmaHakati ngokuvamile agcina phakathi kwamahora ambalwa nezinsuku ezimbili. Leso isikhathi esincane kakhulu sokuthuthukisa umkhiqizo osebenza ngokugcwele. Kodwa-ke, ngasikhathi sinye, kwangiphoqa kakhulu esikhathini esedlule ukubekela phambili izici okufanele ngigxile kuzo. Ngokuvamile, ukuphathwa kwesikhathi kuyikhono elibalulekile lokuba nalo. Lapho unezisombululo eziningana ezingaba khona zokubhekana nenkinga yakho, kufanele ukhethe eyodwa evumelana kangcono nezidingo zebhizinisi ngenkathi uhlonipha izingqinamba zesikhathi.
Okubuye kube kuhle ukuthi ngemuva kwawo wonke ama-hackathon, ungazihlaziya ngokwesikhathi esikuthatha ukuze usebenzise izici ezithile. Isibonelo, ake sithi bekungokokuqala kwadingeka ukuthi uthuthukise ipayipi le-rag, elikuthathe cishe amahora ama-4 ukuwasebenzisa. Ngokuzayo lapho ubhekene nenkinga ye-analogous emsebenzini noma i-hackathon, uzoba nesilinganiso esingcono ngaphambi kwesikhathi esizothatha isikhathi esingakanani uma uthatha isinqumo sokusebenzisa indlela efanayo. Ngaleyo mqondo, isipiliyoni seHachathon sikuvumela ukuthi uchaze kangcono imikhawulo yesikhathi sezindlela ofuna ukuzisebenzisa kumaphrojekthi.
Kimi, isifundo esikhulu kunazo zonke esivela eHackathon besingagxili ekupheleleni lapho sidala i-MVP. Ngenkathi i-MVP ibalulekile, kuyadingeka futhi ukwethula umkhiqizo wakho ukuheha amaklayenti noma abatshalizimali, inkinga iyaxazulula, futhi kungani kungcono kunezixazululo ezikhona emakethe. Mayelana nalokhu, ama-hackathons afundisa ukuthi uqhamuke nemibono engcono exazulula izinkinga zangempela ngenkathi nokuthumela i-MVP ngokushesha, equkethe izici ezibaluleke kakhulu.
Okokufunda abacabanga ngendlela yabo yomsebenzi:I-Roadmap ukuze ube usosayensi wedatha“Uchungechunge lwe-SPans ngokuyisisekelo ngokusebenzisa i-advanced ml. Uma ubuye wabhala kabusha namuhla, iziphi izihloko ezizothuthukiswa, zidilizwe, noma zinqunywe ngokuphelele, futhi ngani?
Ngibhale lolu chungechunge lwendatshana ngonyaka owedlule. Kimina, yonke imiqondo nezihloko engikubhalile zisesikhathini sososayensi bedatha. Zonke izibalo, isayensi yekhompyutha, kanye nezihloko zokufunda ngomshini engilethule kukhona isisekelo esibalulekile kunoma yimuphi unjiniyela wokufunda umshini.
Njengoba manje sesisekupheleni kuka-2025, ngangengeza nesidingo sokuba okungenani nesipiliyoni esincane ngonjiniyela osheshayo kanye nokujwayela amathuluzi athile akhiqizwayo, anjenge-GitHub COPILOT, e-Gemini CLI, kanye nesikhombisi, esizovumela ukusebenza kahle emsebenzini.
Njengenothi, uma kuqhathaniswa neminyaka eyedlule, izinkampani ze-IT zinezidingo eziphakeme kanye nokulindelwe konjiniyela abasencane abangena emkhakheni wesayensi yedatha. Kunengqondo, njengoba amathuluzi anamuhla e-AI angenza kahle imisebenzi ye-junior-level kahle, futhi izinkampani eziningi zithanda ukuthembela kubo manje kunabanjiniyela bokungena, ngoba akudingeki bakhokhe amaholo ngenkathi bethola umphumela ofanayo.
Kungakho-ke, uma umshini wokufunda umshini unawo amakhono aqinile ayisisekelo engiwachazile kulolo chungechunge lwezihloko, kuzoba lula kakhulu kubo ukucwila ngokuzimele ngezihloko eziyinkimbinkimbi.
Isendlalelo sakho sobunjiniyela besoftware kanye ne-ML. Ngabe leso sisekelo sibumba kanjani indlela obhala ngayo?
Ukuba namakhono obunjiniyela obuqinile wesoftware kungenye yezinzuzo ezinhle kakhulu ongaba nazo njengososayensi wedatha:
- Kukwenza uqaphele ukubaluleka kwemibhalo yesoftware ehlelwe kahle futhi wakha amapayipi e-ML ezenziwe kabusha.
- Uqonda kangcono ukuthi ungayenza kanjani ikhodi yakho ihlanzekile futhi ifundeke kwabanye.
- Uqonda izingqinamba ze-algorithmic futhi yiluphi isakhiwo sedatha okufanele ukhethe umsebenzi othile, olususelwa ezidingweni zohlelo.
- Ungasebenzisana kalula nge-BackEnder futhi Deleps Onjiniyela ekuhlanganiseni amamojula wakho wekhodi.
- Awudingi ukuthembela kwabanye ukwenza imibuzo ye-SQL ukuthola idatha edingekayo kusuka database.
Uhlu lungaqhubeka nalo …
Ekhuluma ngezindatshana zami, anginawo abaningi abethula ikhodi eningi. Kodwa-ke, noma nini lapho ngenza, ngilwela ukuyenza ifundeke futhi iqondakale kwabanye. Ngangihlala ngizibeka ezicathulweni zabanye futhi ngizibuze ukuthi umbhalo wami noma izibonelo zekhodi zingaba lula kanjani ukuzibona noma ukuzala uma ngisezicathulweni zabanye. Yilapho isipiliyoni sobunjiniyela besoftware benza lokhu kwangempela kubaluleke kakhulu kimi, futhi ngilandela imikhuba emihle kakhulu yokuhambisa umkhiqizo wami wokugcina.
Ukubheka okwakho iphothifoliyo na- Umuthi othile kikiUhlanganise izisekelo zobunjiniyela besoftware nge-ML kusukela ekuqaleni. Yimuphi umkhuba owodwa wobunjiniyela ofisa ukuthi ososayensi be-data abafisa kakhudlwana bamukela ekuseni?
Onjiniyela abaningi, ikakhulukazi amaJunioors, bavame ukubukela phansi ukubaluleka kokudala imibhalo emihle namapayipi okuvela kabusha. Lokhu kwenzeka kimi esikhathini esidlule, lapho ngigxile kakhulu ekuthuthukiseni amamodeli aqinile noma enze ucwaningo.
Njengoba kwavela, lapho kwadingeka ngiguqule izimo bese kuthi emva kwalokho amasonto ambalwa kamuva ngibuyele emsebenzini owedlule, ngangichitha isikhathi esiningi se-ReadMebook noma ukuhlanganisa zonke izinyathelo ezidingekayo zokuthola zonke izinyathelo ezidingekayo zokusebenzisa amapayipi kusuka ku-zero.
Ngoba bekucishe kube okungenakwenzeka ukufuthelela amapayipi ami ukusuka ekuqaleni, ngangingakwazi ukwenza izivivinyo ngisebenzisa amanye amapharamitha wokungena, okwenza lesi simo sikhungathekise kakhulu.
Kwakuwukuhlangenwe nakho okubuhlungu kimi, kepha esinye sezifundo ezibaluleke kakhulu engizifundile. Ngakho-ke ukube bekufanele nginikeze ucezu lweseluleko kusosayensi wedatha ofisa umkhuba othile, kungaba lokhu:
“Njalo yenza umshini wakho ufunde amapayipi kabusha futhi ubhale kahle”.
Ngonyaka owedlule, ingabe i-AI yashintsha ngokunenhloso indlela osebenza ngayo usuku nosuku njengonjiniyela we-ML? Yini elula, yini eba nzima, futhi yini eyayihlala efanayo?
Eminyakeni yamuva nje, sikubonile ukukhuphuka okukhulu kumathuluzi wobunjiniyela a-AI anamandla:
- I-LLMS, engaphendula cishe kunoma yimuphi umbuzo, nikeza izeluleko, noma ukuthola izimbungulu ku-software
- Isikhombisi, esithandekayo, futhi i-bolt isebenza njengezinto eziphikisayo ze-AI zabathuthukisi
- Ama-Agents AI angaqedela imisebenzi yezinyathelo eziningi
Njengonjiniyela wokufunda umshini, kubalulekile kimi ukuthi ngijwayele nalawa mathuluzi ukuze ngiwasebenzise kahle.
Okwakuba lula
Kusukela ngo-2025, ngiyakwazi ukubuka umthelela omuhle olandelayo emsebenzini wami:
- Kimina, kwaba lula ukuluhlola ngokushesha imibono noma ama-prototypes. Isibonelo, kunezikhathi ezikhona lapho nginikezwa izinkinga zombono zekhompyutha eziye zavela ngaphandle kwendawo yami yolwazi. Ngale ndlela, ngangibuza i-chatgt ukuthi iphakamise imibono eminingana ukuyixazulula. Kwakunezikhathi ezithile lapho i-chatgpt yangikhipha ikhodi, futhi ngazama ukuyikhipha ngaphandle kokuqonda ukuthi isebenza kanjani ngaphakathi.
Ngemuva kwalokho nganginezimo ezimbili ezinokwenzeka:- Uma ikhodi iqhutshwa ngempumelelo futhi ixazulule inkinga yokuqala, khona-ke ngazama ukujula ngaphakathi kwemibhalo ye-OpenCV ukuze ngiqonde ukuthi kusebenza kanjani ekugcineni kanjani.
- Uma ikhodi ayizange uyixazulule inkinga yami, ngingayishaya indiva, bika iphutha ekuxoxeni, noma ngizame ukuthola isixazululo ngokwami.
Njengoba ubona, ngakwazi ukuyovivinya ngokushesha isixazululo esingasebenza futhi ngisindise amahora ocwaningo ngaphandle kwengozi.
- Elinye icala lokusebenzisa elihle kakhulu kimi lalifaka imiyalezo yephutha ngqo ku-chatgt esikhundleni sokufuna isisombululo kwi-Intanethi. Isebenze kahle isikhathi esiningi, kepha kwesinye isikhathi ithinteka ngamaphutha ahlobene nokufakwa komtapo wezincwadi, amaphutha esistimu, kanye nokuhanjiswa kwamapayipi efwini, phakathi kwezinye izindaba.
- Ekugcineni, ngingumuntu omkhulu we-AI Hackathons! Ukuba namathuluzi angakhiqiza i-frontlend kanye ne-backlend yohlelo lwakho kwenza umehluko omkhulu kimi, ngoba manje sengikwazi ukushesha ama-prototypes futhi ngivivinye i-MVP yami ngamahora ambalwa. Engikuthuthukisa manje phakathi nezinsuku zosuku olulodwa kungadinga isonto lonke lomsebenzi.
Okwakuba nzima / kube yingozi
- Lapho ubhala ikhodi nge-AI, kunethuba eliphakeme lokuvuza kwedatha ebucayi. Cabanga nje ukuthi unefayela noma i-Code Fragment equkethe iziqinisekiso ezibalulekile ongazidla ngephutha imodeli ye-AI. Ngemuva kwalokho ithuluzi leqembu lesithathu lizokwazi ubufakazi bakho obucayi. Kungenzeka, ikakhulukazi uma usebenzisa ithuluzi elifana nesikhombisi bese ugcina ubuqiniso bakho kwelinye ifayela kunokuba .Env. Njengomphumela, ngaso sonke isikhathi kudingekile ukuba uqaphele kakhulu.
- Okunye ubungozi akuhloli kahle ikhodi ekhiqizwe yi-AI futhi ungazi ukuthi ungayenza kanjani ukubuyela emuva. Ithuluzi le-AI lingaletha amaphutha angabonakali kwikhodi, ikakhulukazi lapho isetshenziselwa ukuguqula noma ikhodi ekhona yokuphinda i-Reft. Ukuqinisekisa ukuthi ikhodi ekhiqizwe yi-AI ayihlali, kuyadingeka ukubuyekeza kahle izingxenye zekhodi ezikhiqizwayo, zivivinye, bese ugcine ukuguqulwa ngendlela evumela ukuthi uhlale ungakwazi ukubuyela emuva kwindlela edlule, elungile uma kunesidingo.
- Lapho uncika kakhulu kumathuluzi we-AI akhiqizayo kakhulu, kunobungozi ukuthi ikhodi ngeke ifundeke, iqukethe imisebenzi emide ngokweqile, bonisa ukuphindaphinda, noma ukuyeka ukusebenza kahle. Kungakho kubalulekile ukuqonda ukuthi amathuluzi we-AI asebenza ngempumelelo ngokuphumelelayo ekuhlolweni kokugcina kunokugcina ikhodi yokukhiqiza esezingeni eliphakeme.
Okusasele okufanayo
Okusalokhu okubaluleke kakhulu kimi ukuhlakanipha kokuqonda ukuhamba komsebenzi kwangaphakathi kwama-algorithms engikusebenzisayo, ukulondolozela izisekelo zesayensi yekhompyutha, nokubhala ikhodi esezingeni eliphakeme, phakathi kwamanye amakhono asemqoka. Ngamanye amagama, izimiso eziyisisekelo zokuthuthukiswa kwesoftware zizohlala zidingeka ukuze zisebenzise kahle amathuluzi e-AI.
Ngaleyo mqondo, ngithanda ukuqhathanisa isethi yamathuluzi e-AI atholakalayo esikhundleni sonjiniyela osemusha eqenjini lami, engingahumushela kuye imisebenzi emincane. Ngingakucela noma yini engiyifunayo, kepha angikwazi ukuqiniseka ngo-100% kuzokwenza imisebenzi yami kahle, futhi yilapho ukubaluleka kokuthola ubuchwepheshe obuqinile obuqinile kuza.
Ukuze ufunde kabanzi ngomsebenzi kaVyacheslav futhi uhlale usesikhathini ngezindatshana zakhe zakamuva, ungamlandela ku-TDS noma nge-LinkedIn.
