Course Objective & Scope
Bag-of-words methods ignore word order, polysemy and cross-sentence context. This programme teaches neural and modern NLP hands-on: starting from subword tokenisation (BPE/WordPiece), recurrent encoders and BiLSTM-CRF taggers, and moving into training and fine-tuning BERT, RoBERTa, DeBERTa and T5 models with PyTorch and the Hugging Face ecosystem.
Participants build single- and pair-text classification (NLI/entailment), token-level sequence tagging (NER/slot filling), extractive question answering and seq2seq summarisation pipelines on a dataset tailored to their organisation.
Who Should Attend?
- Data scientists, ML engineers and NLP developers
- Teams moving from classical text analytics to deep learning and Hugging Face
- Engineers building in-house classification, summarisation and QA models
- Prerequisite: Python, basic PyTorch and introductory NLP knowledge
Key Takeaways
- Prepare large text datasets with Hugging Face Datasets, fast Tokenizers and dynamic padding
- Fine-tune BERT/DeBERTa encoders for classification, semantic matching and NLI tasks
- Train token-level NER and slot-filling models while preserving subword-to-word label alignment
- Fine-tune encoder-decoder models (T5, BART) for domain summarisation and sequence-to-sequence tasks
- Evaluate neural NLP outputs rigorously using entity-level F1, ROUGE, BERTScore and slice analysis
Curriculum
Neural Language Representations and Sequence Models
- Static vs contextual embeddings; RNN/GRU/BiLSTM encoders and CRF decoding
- Hugging Face Tokenizers, Datasets and dynamic batching with DataCollators
Pretrained Encoder Models: BERT, RoBERTa and DeBERTa
- Masked Language Modeling, [CLS] vs mean pooling and disentangled attention (DeBERTa)
- Fine-tuning for single-text classification and sentence-pair NLI/matching
Token-Level Tasks: NER, Slot Filling and Extractive QA
- BIO/BILOU tagging schemes and subword-to-word label alignment
- Entity-level F1 with seqeval and span-based extractive question answering
Sequence-to-Sequence: Summarisation and Generation with T5/BART
- Encoder-decoder cross-attention and teacher forcing
- Greedy, beam search and nucleus decoding; ROUGE and BERTScore evaluation
Case Implementation and Behavioural Probing
- End-to-end fine-tuning on the enterprise case and baseline comparison
- Inspecting predictions with Integrated Gradients and behavioural checklists
Other Courses in This Category
-
Running NLP models in production at low latency and cost: ONNX/Optimum quantisation, Triton/FastAPI serving, dynamic batching, PII redaction (Presidio), behavioural testing (CheckList) and text drift monitoring.
- Live Course
- 5 modules
-
Turning unstructured documents into structured tables and JSON: hierarchical and multi-label classification, GLiNER and Transformer NER, relation and event extraction, layout-aware PDF parsing and schema-validated extraction.
- Live Course
- 5 modules
-
Enterprise search beyond keyword matching: Sentence-Transformers (bi-encoders), multilingual embeddings, BM25/SPLADE + dense hybrid retrieval, Reciprocal Rank Fusion, Cross-Encoder reranking, Qdrant/pgvector and NDCG@k evaluation.
- Live Course
- 5 modules
-
Deep Learning with PyTorch
Comprehensive deep learning training with PyTorch 2.x covering tensor operations, autograd, CNNs, Transformers, LLM fine-tuning (LoRA/QLoRA), diffusion models, GNNs, and production deployment.
- Live Course
- 12 modules
-
Neural networks, forward pass, backpropagation, activation and loss functions, AdamW, dropout and batch normalisation, giving engineers the skills to build and train their first deep learning models in PyTorch.
- Live Course
- 5 modules
-
Stable, large-scale deep learning covering BF16/FP16 mixed precision, multi-GPU training with DDP/FSDP, self-supervised and contrastive pretraining, metric learning, knowledge distillation and GPU profiling.
- Live Course
- 5 modules
Get in touch
