Alibaba ACA GenAI study plan

Define generative AI core concepts (tokens, embeddings, transformers, inference) in your own words without notes

Explain how Qwen model families differ in scale and modality, mapping each family to a plausible use case

Describe the text-generation request/response lifecycle from prompt to completion

List three common failure modes of LLM outputs and one mitigation for each

Compare prompting techniques (zero-shot, few-shot, chain-of-thought) and pick the right one for three scenarios

Explain how temperature and top-p shape output and predict the effect of changing each

Identify the risks of hallucination and bias and propose a guardrail for each

Summarize how retrieval augmentation improves factual grounding

Navigate Model Studio and locate where models, prompts, and applications are configured

Build a minimal chat application that calls a Qwen model with a system prompt

Explain how API keys, quotas, and rate limits affect application design

Describe how to wire an external knowledge base into an application

Design a multi-step application flow with retrieval and post-processing

Explain how to evaluate application output quality and set an acceptance threshold

Identify where logging and monitoring belong in a deployed application

Draft a cost estimate for a given request volume

Describe the PAI workflow from data preparation through training, evaluation, and deployment

Explain the difference between training, fine-tuning, and inference-serving workloads

Identify which PAI component handles each stage of the lifecycle

List prerequisites for deploying a model as an online endpoint

Explain how to prepare and validate a training dataset for fine-tuning

Describe how to evaluate a fine-tuned model against a baseline

Outline the steps to deploy a model to an endpoint and test it

Identify monitoring signals that indicate a degraded deployment

Describe how an AI coding assistant generates, explains, and refactors code

Explain how repository context and enterprise knowledge sources shape suggestions

Identify security and governance controls for enterprise code assistants

Outline how an enterprise agent is scoped, connected to tools, and monitored

Re-derive the key concept from each domain in a two-minute verbal explanation

Resolve every flagged weak area from earlier self-quizzes

Complete a timed mixed-domain practice set and analyze misses by domain

Build a one-page cross-domain cheat sheet linking concepts to platform features