Ungaba Kanjani Unjiniyela We-AI Ngokushesha (Amakhono, Amaphrojekthi, Umholo)
yindima entsha “eshisayo” endaweni yezobuchwepheshe, futhi abantu abaningi balangazelela ukuthola lo msebenzi.
Ngibona okuthunyelwe okuningi ku-inthanethi okusho ukuthi ungaba kanjani unjiniyela we-AI ezinyangeni ezimbalwa.
Ake ngicacise: noma ngubani okutshela ukuthi ungaba unjiniyela we-AI ezinyangeni eziyisithupha ukudayisela iphupho.
Iqiniso liwukuthi kuzothatha isikhathi eside, kodwa lokho akusho ukuthi awukwazi ukuzama ukusheshisa inqubo.
Uma umusha lapha, ngingu-Egor. Ngisebenza njengonjiniyela wokufunda umshini futhi ngiphinde ngibe umqeqeshi wemisebenzi yabantu abagqekeza kudatha, i-AI, nokufunda ngomshini.
Ngizibonele ngawami ukuthi yini esebenzayo nokuthi yini nje ukumosha isikhathi.
Asingene kukho!
Ake sicacise kahle ukuthi uyini unjiniyela we-AI, njengoba kunokudideka okuningi ku-inthanethi.
Nginesihloko esihlukile esichaza umehluko oyinhloko, kodwa ngamafuphi unjiniyela we-AI ungunjiniyela wesofthiwe ogxile ekusetshenzisweni nasekuhlanganiseni amamodeli ayisisekelo e-GenAI afana noClaude, GPT, BERT, nabanye.
“Abawakhi” lawa mamodeli kusukela ekuqaleni njengososayensi wedatha noma unjiniyela wokufunda ngomshini; kunalokho, bawasebenzisela ukufeza injongo ethile.
Isibonelo, bangase bashumeke i-chatbot kuwebhusayithi yokuthenga ukuze basize amakhasimende athole abakufunayo ngokushesha, noma bengeze umsizi wokubhala amakhodi ku-IDE, njenge-Cursor.
Njengoba onjiniyela be-AI bengonjiniyela besoftware abakhethekile, badinga ukwazi izindlela eziyisisekelo zobunjiniyela besoftware futhi babe nolwazi oluqinile lwezinhlelo ze-AI.
Leli khono lamakhono alivamile kodwa lidingeka kakhulu kulezi zinsuku ngenxa ye-hype ezungeze i-AI. Ngakho-ke ngokwemvelo, iholo lonjiniyela be-AI liphezulu kakhulu futhi izinkampani eziningi zikhokha cishe u-$200k–$300k, ngokuya nge-level.fyi.
Njengoba ubona, kuwumsebenzi omuhle kakhulu onamandla amakhulu okukhula. Manje ake sibheke ukuthi ungaba kanjani munye.
Iqiniso elilodwa elidabukisayo ukuthi kunzima kakhulu ukungena kubunjiniyela be-AI ngaphandle kwesipiliyoni sangaphambilini.
Lokhu kungenxa yokuthi umsebenzi udinga ubuchwepheshe obanele kuyo yonke idatha, ukufundwa komshini, ubunjiniyela besoftware, kanye, ngokwemvelo, ne-AI.
Ngakho-ke, uzodinga ukuba usosayensi wedatha noma unjiniyela wesoftware okungenani unyaka owodwa ngaphambi kokucabanga ukuphendukela ekubeni unjiniyela we-AI.
Ukuthi uba usosayensi wedatha noma unjiniyela wesofthiwe kukuwe kanye nesizinda sakho.
Kodwa-ke, ngincoma ukuthi ngiqale njengonjiniyela wesoftware kuqala, njengoba kuhlobene kakhulu nendima yobunjiniyela be-AI.
Akudingekile futhi ukuthi uthathe izwi lami ngakho; U-Greg Brockman (OpenAI CTO) naye uyavuma ukuthi kungcono ukuba unjiniyela wesofthiwe kuqala bese uthuthukisa ulwazi lwakho lwe-AI/ML.
Njengonjiniyela wesoftware, kufanele wenze umzamo wokufunda amathuluzi nobuchwepheshe obudingekayo ukuze ube unjiniyela we-AI, lokhu kufaka phakathi:
- I-Python – Yonke i-AI/ML ecosystem yakhiwe ngePython, ngakho kufanele ukwazi ukubhala ikhodi yokukhiqiza eqinile ngalolu limi.
- I-SQL –I-AI izungeza idatha, futhi i-SQL iwulimi Lwedatha.
- Amathuluzi Okuthuthukisa Isofthiwe — Udinga ukwazi izinto ezifana gitukulawula inguqulo, zsh/bashizisekelo kanye nokuqonda indlela yokudala nokusebenzisa Ama-API .
- I-System Design Technologies — Uhlelo lwe-AI ozolwakha ekugcineni luzodinga ukukala, futhi cishe uzolusebenzisa ezisekelweni zamafu njenge AWS, I-Azure , I-GCPusebenzisa amathuluzi afana I-Dockerfuthi Kubernetes.
Izinsiza
Umugqa wesikhathi
Umugqa wesikhathi uncike ekutheni kukuthatha isikhathi esingakanani ukuthola umsebenzi wobunjiniyela besofthiwe noma wesayensi yedatha.
Ukuba ngokoqobo, uma unesizinda se-STEM esinolwazi oluthile oluqinile, futhi uzikhandla ngempela, ungakwazi ukuthola imisebenzi kulezi zindima cishe ezinyangeni eziyisi-6.
Kufanele-ke uhlale kule ndima isikhathi esingangonyaka ngaphambi kokuzama ukushintshela kubunjiniyela be-AI ukuze uqinisekise ukuthi uhlanganise izinto eziyisisekelo.
Kunemihlahlandlela eminingi eku-inthanethi yokuthi ungangena kanjani kubunjiniyela besofthiwe, futhi nginemigwaqo eminingana yokuba usosayensi wedatha ongayihlola.
Ngokuhambisana nomsebenzi wakho wesikhathi esigcwele njengonjiniyela wesofthiwe, uzodinga ukuzithuthukisa wena ezisekelweni ze-AI/ML ukuze uqinisekise ukuthi uthuthuka ngokushesha ohambweni lwakho.
Ngokuqinisekile awudingi ukuba ne-PhD ezingeni lokuqonda leMaths, njengoba ungeke uwakhe la mamodeli kusukela ekuqaleni, kodwa izokunikeza imininingwane engemuva ukuze ungene ujule ezihlokweni ezithuthuke kakhulu ngokuhamba kwesikhathi.
Lezi yizinto okufanele uzazi:
- Izibalo Eziyisisekelo –Ukubuka konke okuqinile kwezibalo, amathuba, i-algebra yomugqa kanye nokubala kuzokusiza uqonde ukuthi kwenzekani ngaphansi kwehood.
- Ukufunda Okugadiwe –Yazi ukuthi ama-algorithms ayisisekelo afana nokuhlehla komugqa, izihlahla zezinqumo nemishini yokusekela ye-vector asebenza kanjani.
- Ukufunda Okungagadiwe –Yazi ukuthi ama-algorithms ayisisekelo afana ne-K-Means kanye ne-K-Nearest-Neighbour asebenza kanjani.
- Amanethiwekhi Neural —Lezi zingumgogodla wama-LLM, futhi ukuqonda kahle izihloko ezifana nokusabalalisa okusemuva, ama-gradient ashabalalayo nemisebenzi yokuvula kuzokuvumela ukuthi ulungise amamodeli e-AI ngokushesha ngokuzayo.
- Izisekelo zama-LLM –Noma ungeke uwakhe ama-LLM kusukela ekuqaleni, uzobe usebenza nawo nsuku zonke, ngakho-ke kuhle ukuba nolwazi oluthile mayelana nendlela asebenza ngayo. Kufanele ufunde mayelana nezindawo ezinjengama-transformer, ama-autoencoder, amathokheni, nokushumeka.
Izinsiza
Umugqa wesikhathi
Ukufunda okuyisisekelo kuzoncika ekutheni ufunda isikhathi esingakanani ngenkathi usebenza njengonjiniyela wedatha/unjiniyela wesoftware.
Isincomo siwukuhlanganisa le mibono emsebenzini wakho wansuku zonke ngangokunokwenzeka.
Ukube bengifunda konke lokhu ngaphandle kwamahora okusebenza, bengingalindela ukuthi kuzothatha izinyanga ezi-3–6 uma uzikhandla.
Kuleli qophelo, sekuyisikhathi sokungena ujule emibonweni ethile kanye nemibono ozobe uyisebenzisa njengonjiniyela we-AI emhlabeni wangempela.
Lo mkhakha uthuthuka ngokushesha, futhi nyanga zonke kuba “nento” entsha okufanele ifundwe. Ngizobala izinto eziyisisekelo ezingaphelelwa yisikhathi lapha njengoba zibaluleke kakhulu.
- Ama-API we-AI– Izinkonzo ezifana I-OpenAI's API ikuvumela ukuthi uhlanganise amamodeli anamandla ngaphandle kokudinga ukuzakhela ngokwakho. Lena indlela eshesha kakhulu yokuqala ukwakha izinhlelo zokusebenza zangempela ngamakhono e-AI.
- Ubunjiniyela obusheshayo– Ukufunda ukuxhumana ngempumelelo namamodeli e-AI kuyikhono elibalulekile. Ukwaziswa okuklanywe kahle kungathuthukisa kakhulu imiphumela yamamodeli futhi kubalulekile ukuze uthole imiphumela engaguquki.
- Ukubuyisa Isizukulwane Esingeziwe (RAG)– Qonda ukuthi ungaxhuma kanjani kuma-LLM kusizindalwazi sangaphandle njenge Iphayinifuthi usebenzise ulwazi oluhlobene ukuze uthuthukise ukunemba kwezimpendulo zemodeli ye-AI.
- I-Model Context Protocol (MCP) – Indlela ejwayelekile yokuxhuma amamodeli akho e-AI kuzinhlelo zokusebenza zangaphandle ezifana namafayela, amaseva nezinye izinhlelo zokusebenza.
- I-LangChain–Leli yiphakheji elihle kakhulu lokusebenza namamodeli we-AI ePython. Inikeza zonke izakhiwo ozidingayo ukuze wakhe futhi uxhume ama-LLM ngaphandle komthungo.
- Ukuhleleka Kwayo–Qonda ukuthi ungathuthukisa kanjani ukusebenza kwemodeli ye-AI ngokuyiqeqesha kudatha ethile ukuze ibe ngcono ekuphenduleni nasekunikezeni imiphumela yecala elithile lokusetshenziswa.
Izinsiza
Umugqa wesikhathi
Ukufunda le miqondo kuzothatha isikhathi esincane kancane kunokufunda okuyisisekelo kwe-AI/ML, njengoba kuncane okungenziwa.
Ngingalindela ukuthi kuzothatha cishe izinyanga ezingu-2-3 ukufunda yonke into ibe sezingeni elihle.
Kuningi ukudideka kokuthi yimaphi amaphrojekthi okufanele uwakhe ukuze uthole umsebenzi kubunjiniyela be-AI
Ukukubeka kalula, amaphrojekthi amahle kakhulu yilawo akukhuthazayo futhi azuzisa uhlobo oluthile lomsebenzisi wokugcina noma iklayenti.
Nazi izinyathelo zezinga eliphezulu:
- Umbono– Xoxa ngemibono nezihloko eziqondene nawe kanye nenkinga ofuna ukuyixazulula. Lokhu kufanele kuvele emicabangweni yakho kanye nocwaningo lwakho; ungabheki ku-inthanethi noma ucele imibono yephrojekthi kubantu abafana nami. Noma yini engikunika yona izoba iphrojekthi embi kuwe ngokushesha.
- Idatha– Thola idatha yenoveli nejabulisayo usebenzisa ama-API omphakathi, amawebhusayithi kahulumeni, ukuklwebheka kwewebhu, njll. Ufuna ukuphindaphinda idatha engcolile ozohlangana nayo emhlabeni wangempela.
- Sebenzisa– Udinga ukukhombisa ikhono lakho lokuphakela amasistimu e-AI ekupheleni ukuya ekupheleni. Lokhu kuzobandakanya ukugcinwa kwedatha, ukuhlanzwa kwedatha, ukuxhumeka kwemodeli bese kuba nokuhlanganiswa okuthile ngaphambili nge-API noma uhlelo lokusebenza lwewebhu. Udinga ukufanisa umsebenzi ozobe uwenza njengonjiniyela we-AI wesikhathi esigcwele ngokuseduze ngangokunokwenzeka.
- Idokhumenti— Akekho ozokwazi ngephrojekthi yakho uma ungatsheli abantu ngayo. Yenza okuthunyelwe kwe-LinkedIn, bhala indatshana yebhulogi bese uyengeza kuphothifoliyo yakho. Qiniseka ukuthi iphrojekthi yakho ine-README ecacile, ehlelwe kahle ku-GitHub ukuze abantu bazihlolele yona. Yabelana ngomsebenzi wakho ngangokunokwenzeka, njengoba kuzokwandisa amathuba akho okubonwa abantu abangaba abaqashi.
Umugqa wesikhathi
Ukudala amaphrojekthi amahle kanye nokwakha iphothifoliyo eqinile kuzothatha isikhathi. Ngokufanelekile, kufanele wakhe amaphrojekthi amabili aphezulu kufanele akuthathe cishe izinyanga ezi-3 sezizonke. Lokhu kucabanga ukuthi unganikela ihora elingu-1 ngosuku ukwakha lezi.
Lokhu kungaba okuthunyelwe kukodwa, kodwa ake ngikunikeze izinga eliphezulu elingu-80/20 lalokho okufanele ukwenze:
Qalisa kabusha
Ukuze ubhale ukuqhubeka okuhle, qiniseka ukuthi yonke into imayelana nobunjiniyela be-AI:
- Yiba namakhono akho obuchwepheshe phezulu ngamathuluzi afanelekile nobuchwepheshe bezindima zobunjiniyela be-AI.
- Yenza amaphrojekthi akho abonakale ngokucacile ngamamethrikhi, izibalo, futhi, ikakhulukazi, umthelela wezezimali.
- Kugcine kulula: imibala emaphakathi, ikholomu eyodwa, amafonti afundeka kalula, kanye nekhasi elide kuphela.
- Faka ohlwini ulwazi lwakho olufanele njengonjiniyela wesofthiwe noma usosayensi wedatha.
Nginesihloko esigcwele sendlela yokwenza i-resume enhle ongayibheka ngezansi, kanye nesifanekiso esenziwe ngomumo ongasisebenzisa.
I-LinkedIn
Yenza iphrofayili yakho ye-LinkedIn icace ukuthi ufuna izindima zobunjiniyela be-AI:
- Isihloko sakho kufanele siqukathe “AI Engineer”, hhayi “sifisa” sicela. Isibonelo, ubani ongafuna ukuqasha udokotela wamazinyo “ofisa”?
- Faka amagama angukhiye kuzo zonke izigaba zakho zokuthi “Mayelana nami” kanye “Nesipiliyoni”, kodwa wengeze ngokwasemzimbeni futhi ungabhali izigaba.
- Yenza iphrofayela yakho ibe yinhle ngesithombe esicacile nesibhengezo esibukeka kahle. Lokhu kwenza umehluko omkhulu kunalokho ocabanga ngakho.
Ukudluliselwa Nokuxhumana
Abantu abaningi bacabanga ukuthi badinga ukwakha inqwaba yamaphrojekthi futhi bathathe izifundo ezingapheli ukuze bavelele futhi bathole umsebenzi.
Lokho kuwukuchitha isikhathi ngokuphelele.
Ukudluliselwa kuyithikithi eligolide lanoma yimuphi umsebenzi wezobuchwepheshe.
Ngokocwaningo, ukudluliselwa kwezicelo kubalelwa ku-7% wezicelo kodwa u-40% wabo bonke abaqashwayo. Uma uthunyelwe, cishe unethuba eliphindwe ka-6 lokuthola umsebenzi wamaphupho akho.
Leyo mandla iyahlanya.
Indlela othola ngayo ukudluliselwa empeleni ilula, futhi okudingayo wukuzethemba okuthile ohlangothini lwakho.
- Thola izinkampani eziqashela onjiniyela be-AI noma izinkampani ongathanda ukuzisebenzela.
- Phequlula abasebenzi babo ku-LinkedIn futhi uthole othile ofana nawe. Lona kungaba umuntu onenyuvesi efanayo nesizinda, unjiniyela we-AI futhi.
- Xhuma futhi ubathumele i-DM equkethe okuthile okuthandile mayelana nephrofayela yabo, uhambo nanoma yini yomuntu siqu. Ungalokothi ucele ukudluliselwa kumlayezo wokuqala.
- Xoxa nabo futhi ubabuze imibuzo ngomsebenzi wabo, amaphrojekthi nanoma yini enhle abayenzayo.
- Ngemva kwemilayezo embalwa, yilapho ucela khona ukudluliselwa noma iyiphi impendulo ku-resume yakho.
Inqubo ilula kakhulu, inkinga abantu bayesaba kakhulu ukuyenza.
Kodwa-ke, angikaze ngibe nesipiliyoni esibi, ngoba uhlale uhola ngokuncoma noma ukuvula ngabo.
Abantu bayakuthanda ukukhuluma ngabo, futhi okudingeka ukwenze nje ukuzitholela ubungane futhi ubonise ukuthi unentshisekelo kubo.
Umugqa wesikhathi
Ukuthola umsebenzi kungahluka kakhulu, futhi kungase kube nenhlanhla ngezinye izikhathi. Kodwa-ke, ngokulandela ukudluliselwa nokugwema iziphazamiso zamaphrojekthi nezifundo, lokhu kufanele kuthathe izinyanga eziyisi-6.
Ngakho-ke, ukuze ube unjiniyela we-AI, kuzokuthatha, ngokunethemba, cishe iminyaka emi-2, kodwa futhi udinga ukuthola umsebenzi njengonjiniyela wesoftware noma usosayensi wedatha kuqala.
Lokhu kungase kubonakale kuyisikhathi eside, kodwa lezi zindima zinamakhono aphezulu futhi zikhokha amaholo angenangqondo. Ngeke ulindele ukwenza izifundo ezimbalwa bese ungena uqonde kuzo.
Uma ngemva kokufunda lesi sihloko, ufuna ngempela ukuba unjiniyela we-AI, kuhle lokho!
Kodwa-ke, njengoba ngisanda kusho, udinga ukuba usosayensi wedatha kuqala. Ngenhlanhla, kwesinye sezindatshana zami ezedlule, ngibhale ncamashi izinyathelo engizozilandela uma ngiphinde ngibe usosayensi wedatha.
Ngizokubona lapho!
Joyina incwadi yami yezindaba yamahhala lapho ngabelana ngamathiphu amasonto onke, imininingwane, nezeluleko ezivela kokuhlangenwe nakho kwami njengososayensi wedatha oprakthiza kanye nonjiniyela wokufunda ngomshini. Futhi, njengomuntu obhalisile, uzothola okwami Isifanekiso SAMAHHALA Sokuqalisa kabusha!
Ukukhipha Idatha
Ama-imeyili amasonto onke akusiza ukuthi uthole umsebenzi wakho wokuqala kusayensi yedatha noma ukufunda ngomshininewsletter.egorhowell.com
