This 2025 “Heima Programmer Python AI Development” mega-course is a 1000+ hour, industry-oriented learning path that takes you from zero to production-ready AI engineer. The curriculum covers the complete life-cycle: Linux & Python fundamentals, data engineering, classical ML, modern deep learning, large-language models, computer vision, NLP, knowledge graphs, financial risk control and MLOps deployment.
Highlights of the 14 stages:
① Python basics & advanced tricks, PEP8, debugging, testing, packaging;
② Linux command-line, MySQL relational modeling, SQL tuning, indexing;
③ NumPy vectorization, Pandas data wrangling, Matplotlib/Seaborn visualization, 32 real-world cases (retail, e-commerce, weather, social media);
④ 10 classic ML algorithms—KNN, linear & logistic regression, decision trees, random forest, Adaboost, XGBoost, naive Bayes, SVM, K-means, PCA—plus feature engineering, hyper-parameter search with Optuna, cross-validation and model explainability;
⑤ Deep-learning crash course: PyTorch tensors, autograd, CNN, RNN, LSTM, GRU, attention, Transformer, BERT, GPT, image classification, text generation and time-series forecasting;
⑥ Computer vision 68 lectures: OpenCV image processing, geometric & photometric transformations, YOLOv8 object detection, face detection, alignment, recognition and payment system, lane-line detection, vehicle counting, DeepSort multi-object tracking and Kalman filtering;
⑦ NLP 13 days: jieba & HuggingFace tokenizers, Word2Vec, FastText, RNN, Bi-LSTM, GRU, Seq2Seq, Bahdanau & self-attention, Transformer, BERT, GPT, prompt engineering, LoRA, prefix tuning, RLHF, LangChain, vector DB, RAG and conversational agents;
⑧ Knowledge graph 10 days: named-entity recognition with LSTM-CRF, relation extraction with CasRel, Neo4j graph storage, Cypher query and visualization;
⑨ Large-language-model 13 days: theory and hands-on for LLaMA, Bloom, ChatGLM, GPT-3/4, data parallel, model parallel, DeepSpeed ZeRO, CPU/GPU mixed precision, parameter-efficient fine-tuning (LoRA/QLoRA, P-Tuning v2), evaluation metrics (Perplexity, BLEU, ROUGE), cloud GPU practice, prompt optimization, code generation and document extraction;
⑩ Financial risk-control 8 days: credit business, ABC scorecards, rule mining, feature derivation, time-window splits, ChiMerge binning, WOE/IV, logistic regression scorecard, LightGBM ensemble, class-imbalance handling (SMOTE, cost-sensitive), anomaly & fraud detection and model monitoring;
⑪ Capstone projects: text classification for investment advice, face-payment system, AI smart-traffic (lane line + vehicle counting), RFM customer segmentation, retail-member data dashboard, stock quantitative back-test, knowledge-graph QA and local LLM knowledge base;
⑫ MLOps & deployment: Git, Docker, Kubernetes, CI/CD, GPU multi-node training, ONNX export, TensorRT/OpenVINO acceleration, REST & gRPC serving, Prometheus+Grafana monitoring;
⑬ Interview preparation: 300 leet-code algorithms, system-design questions, ML theory Q&A, resume templates and mock interviews.
All datasets, slides, codes and pre-trained models are open-source. After graduation you will be qualified for positions such as Algorithm Engineer, Computer-Vision Engineer, NLP Engineer, LLM Application Architect, Financial Modeler or Data Science Consultant.
这份 2025 版「黑马程序员 Python 人工智能开发」合集是一套面向就业与科研的超大体系课程，总课时 1000+，源码、笔记、数据全部开源。内容从 Python 语法、Linux 运维、MySQL 数据库起步，贯穿数据采集、清洗、可视化、统计分析、机器学习、深度学习、大模型，直至工业级部署与上线监控，形成“数据→算法→模型→产品”完整闭环。
课程分为 14 大阶段：①Python 基础与高级编程；②面向对象、并发、网络编程；③文件操作、异常、模块与包；④MySQL 数据库与 SQL 优化；⑤NumPy、Pandas、Matplotlib、Seaborn 数据分析；⑥32 个实战案例涵盖零售、电商、金融、气象、社交媒体；⑦机器学习 10 大算法：KNN、线性回归、逻辑回归、决策树、集成学习（随机森林、Adaboost、XGBoost）、朴素贝叶斯、SVM、K-Means、PCA、特征工程与超参优化；⑧深度学习 6 天速成：PyTorch 张量、自动微分、CNN、RNN、LSTM、GRU、注意力、Transformer、BERT、GPT、图片分类、文本生成、时间序列预测；⑨计算机视觉 68 讲：OpenCV 图像处理、目标检测（YOLOv8）、人脸检测、人脸识别、人脸支付、车道线检测、车流量统计、多目标跟踪、DeepSort、卡尔曼滤波；⑩自然语言处理 13 天：分词、词向量、Word2Vec、FastText、RNN、LSTM、GRU、Seq2Seq、Attention、Transformer、BERT、GPT、预训练+微调、Prompt Engineering、LoRA、Prefix Tuning、RLHF、LangChain、向量数据库、RAG、Agent、中文文本分类、情感分析、机器翻译、问答系统；⑪知识图谱 10 天：实体识别、关系抽取、LSTM+CRF、CasRel、Neo4j 图数据库、Cypher 查询、可视化；⑫大模型 13 天：LLaMA、Bloom、ChatGLM、GPT-3/4 原理、并行训练、DeepSpeed、ZeRO、LoRA/QLoRA 微调、P-Tuning v2、指令跟随、人类反馈强化学习、评估指标（Perplexity、BLEU、ROUGE）、趋动云算力实战、Prompt 优化、对话系统、代码生成、文档抽取、开源大模型平台（讯飞星火、百度千帆、阿里百炼、PAI）；⑬金融风控 8 天：信贷业务、ABC 卡、规则挖掘、特征衍生、时间切片、卡方分箱、WOE/IV、逻辑回归评分卡、LightGBM 集成评分卡、样本不平衡、SMOTE、代价敏感、异常检测、反欺诈、模型监控；⑭亿级项目实战：投满分文本分类、人脸支付、AI 智慧交通（车道线+车流量）、RFM 用户画像、零售会员数据可视化、股票量化回测、知识图谱问答、大模型本地知识库、GPU 分布式部署、Docker+K8s 上线、CI/CD 流程。
课程特色：
工业真数据：金融、电商、交通、医疗、政务等 50+ TB 脱敏数据；
全链路 MLOps：数据版本管理→特征仓库→模型注册→灰度发布→线上监控；
多语言混合：Python、C++、CUDA、SQL、Shell、YAML、Dockerfile；
多框架协同：PyTorch、TensorFlow、Paddle、MindSpore、ONNX、TensorRT、OpenVINO；
就业导向：简历模板、面试题库、算法手撕、HR 面指导、内推通道。
结业可胜任算法工程师、CV/NLP 工程师、大模型应用架构师、金融风控建模师、数据分析专家等岗位。
Информация по комментариям в разработке