Skip to main content

Yeh Kitab Jin Roles Ki Training Deti Hai

Market jitni tezi se naye titles ijaad kar rahi hai, utni tezi se unhein define nahin kar pa rahi. In mein se zyada tar titles asal mein aik hi discipline hain, bas mukhtalif depths par: wohi discipline jo yeh kitab sikhati hai. Yahan us ka map hai, aur theek theek yeh ke kitab har role ki taraf aap ko kitni door le jati hai.

Yahan hum naye agentic AI daur ke roles define karte hain - woh jobs jo is liye wujood mein aayi hain ke companies ab AI Workers manufacture, run, aur govern karti hain. Entries is hisaab se sort ki gayi hain ke kaam asal mein kis tarah cluster hota hai, aur har aik ke saath di gayi verdict honest scope line hai: kitab aap ko us ki taraf kitni door le jati hai, aur kahan se certification tracks sambhal leti hain. Verdicts naam se zyada matter karti hain. Jahan kitab rukti hai, woh yeh keh deti hai.

Kitab general agent use ke dono modes ki training deti hai. Mode 1 general agent ko apna kaam tez karne ke liye use karna hai - aik proficiency jo har reader ko chahiye, koi job title nahin. Mode 2 aise AI Workers manufacture karna hai jo kaam aap ke liye karte hain, aur job titles wahin rehte hain. Neeche diya gaya map Mode 1 baseline se khulta hai, phir Mode 2 roles ki taraf jata hai - jo lag bhag poora map hi hain.

Is vocabulary mein naye hain (Digital FTE, SKILL.md, Agent Factory)? Pehle Thesis aur Glossary parhein; yeh page in ko maan kar chalta hai.

Role map: aik core pipeline, woh roles jo isay extend aur support karte hain, deliberate stops, aur Mode 1 baseline

Poora map aik nazar mein - core pipeline, kya isay extend aur support karta hai, kitab kahan rukti hai, aur in sab ke neeche baseline.

Woh baseline jahan se sab shuru karte hain​

Mode 1 Practitioner - koi title nahin, aik proficiency. Koi Worker manufacture karne se bahut pehle, aap general agent ko apna kaam tez karne ke liye use karte hain: reason karne, likhne, code karne, analyze karne, plan karne, aik outcome ship karne, aur session close karne ke liye. Yeh Mode 1 hai, aur kitab isay sab ke liye sikhati hai - engineers ko Claude Code ya OpenCode ke zariye, domain experts ko Claude Cowork ya OpenWork ke zariye - Seven Principles of General Agent Problem Solving ke under. Yeh woh on-ramp hai jis par har reader neeche diye gaye Mode 2 roles se pehle chalta hai, aur yeh aap ko aik naya kaam dene ke bajaye us kaam mein sharper banata hai jo aap pehle se kar rahe hain. Woh floor jis par sab khare hote hain, koi title nahin jo aap rakhte hon.

Generalist core​

Yeh core roles aik hi pipeline ki tarah chalte hain, intent se production tak: Outcome Architect (kya) → Digital FTE Builder (banao) → AI-Native Company Architect (system) → Cloud AI Engineer (chalao). Isay apni company ke andar chalao to yeh chaar roles hain; kisi client ki company ke andar chalao, aur aik embedded, vendor-neutral engineer isay end to end carry kare, to yeh Forward Deployed Engineer hai. Map par baqi sab kuch is line ko support, extend, ya bound karta hai.

Core pipeline: chaar internal roles, ya client par aik embedded Forward Deployed Engineer

Chaar roles yeh line aap ki apni company ke andar chalate hain; aik embedded engineer wohi line client ke andar carry karta hai.

Outcome Architect: intent own karta hai, execution nahin. Agent daur mein kaam teen tarah se split hota hai - intent, execution, verification. Worker execution own karta hai; yeh role intent own karta hai. Yeh decide karta hai ke Worker ko kya achieve karna chahiye, woh spec likhta hai jo isay pin karti hai, set karta hai ke "correct" ka matlab kya hai, aur priority deta hai ke kaun se Workers banaye bhi jayein - woh insan jo kya aur kyun ka jawab Builder ke kaise batane se pehle deta hai. Jahan Strategist track client-facing discovery aur ROI own karti hai, wahan Outcome Architect internal Worker roadmap aur us ke peeche ki specs own karta hai. Kitab is ki seedhi training deti hai: spec-driven development apni jarh mein wohi discipline hai - aisa intent likhna jis par Worker ko accountable rakha ja sake. Training deti hai - woh discipline jis par poora method tikta hai.

Digital FTE Builder: unit product, end to end banaya hua. Market isay AI Engineer kehti hai - us ka catch-all naam us shakhs ke liye jo AI components se applications banata hai aur AI coding agents chalata hai. Is kitab ka naam sharper hai, kyun ke jo cheez aap banate hain woh sharper hai: Digital FTE, woh unit jis se poori company assemble hoti hai. Yeh kitab ka primary graduate hai. Yeh poori spine ki training deti hai: spec-driven development, SKILL.md authoring, agent architecture, tool aur MCP interfaces, evaluation, aur human oversight - itni deployment ke saath ke ship kiya ja sake, aur baqi depth Cloud AI Engineer par chhori. Training deti hai, end to end.

AI-Native Company Architect: company design karta hai, single Worker nahin. Poora enterprise - Two-Layer Model, management layer, workforce, woh nervous system jo un ke darmiyan events carry karta hai, aur woh system of record jis ke against yeh sab chalta hai. Agent Factory woh process hai jo yeh architect practice karta hai; AI-Native Company woh product hai jo yeh ship karta hai. Kitab is ka canonical source hai. Paanch-quarter ka Certified Agentic AI Architect program is ka credential hai. Poori training; Architect track se certified.

Cloud AI Engineer: woh jo AI Worker aur AI-Native company ko production mein chalata hai. Digital FTE banana kaam ka aik hissa hai; usay reliably chalana doosra hai - aur isi tarah poori AI-Native company chalana jis ka woh hissa hai. Jahan AI-Native Company Architect enterprise design karta hai, wahan yeh role isay operate karta hai: Workers, management layer, aur nervous system ko real cloud infrastructure par deploy aur scale karta hua - ship karne ke liye Azure Container Apps, durable execution ke liye Inngest, scale ke liye Dapr aur Kubernetes. Yahin system prototype rehna chhor kar aisi company ban jata hai jis par koi organization bharosa kar sake. Training deti hai, end to end.

Do roles jin ki training sirf yeh kitab deti hai​

Subject Matter Expert as Skill Author: woh role jisay market ne abhi tak naam nahin diya. Woh accountant, lawyer, ya supply-chain expert jo apna judgment SKILL.md mein encode karta hai aur Digital FTE ka knowledge engine ban jata hai. Zyada tar market lists is role ko miss karti hain kyun ke woh ab bhi AI work ko sirf engineering samajhti hain. Yeh kitab domain judgment ko khud aik aisi cheez maanti hai jise author, test, aur deploy kiya jata hai. Yeh do mein se aik role hai jin ki training sirf yeh kitab deti hai. Poori training deti hai: judgment andar, kaam karta hua agent bahar.

Forward Deployed Engineer (FDE): vendor-neutral version jo market ko nahin milta. Palantir ne FDE ki bunyaad rakhi; AI vendors abhi isay revive kar rahe hain, engineers ko client organizations ke andar embed kar ke taake agentic workflows ko client ki reality ke saath fit kiya ja sake. Woh technical half bilkul Agent Factory work hai, aur kitab is ki poori training deti hai. Twist vendor neutrality hai. Jaisa ke Andrew Ng ne The Batch mein note kiya, clients ko aise FDEs dhoondhne mein dushwari hoti hai jo kisi aik vendor se bandhe na hon - yeh role is liye exist karta hai ke aik vendor ki product company mein gehrai tak wire kar de. Is se baad mein switch karne ki freedom kam ho jati hai. Yahan ka method kisi vendor se bandha nahin, is liye kitab woh FDE produce karti hai jo market ko bar bar nahin milta, us optionality ke saath intact. Yeh doosra role hai jis ki training sirf yeh kitab deti hai. Job ka doosra half Certified Agentic AI Business Strategist track ka hai, core kitab ka nahin: client discovery, prioritization, ROI framing, aur woh discipline jo aik ghair-haqeeqi maang ko push back kar sake. Technical core ki training deti hai; consulting layer Strategist track mein rehti hai.

Supporting roles​

Evals Engineer: verification specialist. Verification yahan koi baad ki cheez nahin; yeh survival standard hai. Core curriculum, koi add-on nahin.

AI Governance Officer: woh rules likhta hai jin ke under workforce chalti hai. Har AI Worker aik authority envelope ke andar operate karta hai - woh kya akela decide kar sakta hai, kya escalate hona chahiye, kya woh kabhi touch nahin kar sakta. Yeh role woh envelope institutional level par likhta hai: risk tiers, escalation aur approval policy, audit aur liability rules, aur woh mapping jo company ko jis cheez ka jawab dena hai us se jorti hai - bank mein model risk aur fair lending, kahin aur data residency. AI-Native Company Architect woh mechanism banata hai jo aik envelope enforce karta hai; Governance Officer decide karta hai ke us mein kya likha ho. Digital FTE Supervisor aik deployed Worker ka jawab deta hai; Governance Officer woh rules set karta hai jin ke andar har Worker aur supervisor kaam karta hai. Kitab is framework discipline ki seedhi training deti hai; aap ki industry ke specific regulations woh inputs hain jo aap laate hain. Governance framework ki training deti hai; aap ke jurisdiction ke rules aap ne supply karne hain.

Digital FTE Supervisor: accountable insan, Worker dar Worker. Woh insan jo aik deployed AI worker ki accountability own karta hai: human-in-the-loop, reviewer, woh naam jis ki taraf audit trail ishara karta hai. Yeh operator ka kaam title bana hua hai - kaam chalana, worker banana nahin. Training deti hai.

Jahan kitab jaan boojh kar rukti hai​

LLMOps Engineer: model tak, model khud tak nahin. Agents ko production mein chalana Cloud AI Engineer ka kaam hai, aur kitab is ki training deti hai. Kitab fine-tuning ki bhi hands-on training deti hai - lekin last resort ke taur par, default ke taur par nahin. Aik fine-tune aap ke system ko aik model snapshot se baandh deta hai aur us optionality ki qeemat leta hai jise poora method protect karta hai, is liye aap isay tabhi reach karte hain jab prompting, context, tools, aur retrieval waqai kam par jayein. Hard stop model khud banana hai: aik foundation model ko scratch se pre-train karna scope se bahar rehta hai, kyun ke woh capability commoditize ho rahi hai. Fine-tuning aur model ke gird ops ki training deti hai, foundation models banane ki nahin.

Harness Engineer: woh runtime jo aap use karte hain, woh nahin jo aap banate hain. Harness agent runtime hai - OpenAI Agents SDK, Claude ke managed agents, aur is qisam ke - jo agent loop chalata hai, state manage karta hai, aur tool calls execute karta hai. Kitab aap ko in ko fluently use karna aur un ke across portable rehna sikhati hai, kyun ke aap ki discipline us runtime se zyada jeeti hai jo bhi jeete. Runtime khud banana yeh job nahin. Us operator ki training deti hai jo koi bhi runtime use karta hai, us engineer ki nahin jo runtime banata hai.

AI Data Engineer: agent-facing data layer. System-of-record ka kaam agent-facing data layer ko touch karta hai: Postgres, pgvector, aur MCP woh spine ke taur par jis se agent read karta hai. Classic pipeline aur warehouse engineering adjacent hai, central nahin. Agent-facing data layer ki training deti hai, general data engineering ki nahin.

Pattern hi asal nishani hai. Agent daur kaam ko bahut se roles mein phaila deta hai, aik mein nahin - Workers banana, unhein chalana aur govern karna, unhein judgment sikhana. Map hi asal baat hai: dhoondhein ke aap pehle se kahan khare hain, aur kitab aap ko wahan se kitni door le jati hai.