What a Junior } should know

Knows Elasticsearch basics: indexes, documents, basic queries. Performs simple full-text search queries for NLP tasks. Understands analyzer and tokenizer concepts.

Knows pandas basics for working with text data: loading corpora, filtering, grouping, basic text preprocessing. Uses str accessor for text column operations.

Knows scikit-learn basics for NLP: TF-IDF vectorizer, text classification via SVM/Naive Bayes, Pipeline. Trains baseline NLP models and evaluates via cross-validation.

Knows experiment tracking basics: MLflow, Weights & Biases. Logs NLP experiment metrics and parameters: F1, precision, recall for NER and text classification models.

Knows basics of LLM applications for NLP tasks: text generation, summarization, few-shot classification. Uses Hugging Face transformers for basic text processing tasks.

Knows ML pipeline basics for NLP: data collection, preprocessing, feature extraction, training, evaluation. Uses scikit-learn Pipeline and spaCy for building simple NLP pipelines.

Knows MLflow basics: experiments, runs, metrics, parameters, artifacts. Logs NLP model training results: F1-score, accuracy, confusion matrix for text classification and NER.

Knows basics of NLP model monitoring: quality metrics, data drift, concept drift. Sets up basic alerts for metric degradation of text classification and NER models in production.

Knows NLP model serving basics: REST API endpoints, model loading, batching. Deploys simple NLP models as REST API for text classification and NER tasks.

Knows NER basics: entity types (PER, ORG, LOC), BIO tagging, basic approaches. Applies pre-trained spaCy NER models and evaluates quality via F1-score.

Knows PyTorch basics for NLP: tensors, autograd, nn.Module. Trains simple NLP models: text classification via LSTM, embedding layers for word representations. Understands training loop.

Knows RAG basics: retrieval, augmentation, generation. Applies simple RAG pipelines for NLP tasks: searching relevant documents and generating context-based answers.

Knows sentiment analysis basics: polarity, subjectivity, aspect-based approaches. Applies pre-trained models for sentiment detection: VADER, TextBlob, Hugging Face sentiment pipeline.

Knows text classification basics: bag-of-words, TF-IDF, basic classifiers. Trains simple models for text categorization, spam filtering. Evaluates via accuracy, F1-score.

Knows transformer architecture basics for NLP: self-attention, positional encoding, BERT/GPT. Uses Hugging Face transformers for inference: text classification, NER, summarization.

Knows vector database basics: embeddings, similarity search, ANN algorithms. Uses Pinecone/Weaviate/Qdrant for storing text embeddings and semantic search.

Knows core ChatGPT and Claude API capabilities: request formats, generation parameters. Uses LLM API for simple NLP tasks: summarization, entity extraction from text.

Knows prompt engineering basics for NLP tasks: role prompting, few-shot examples, chain-of-thought. Crafts prompts for entity extraction, text classification, and summarization.