I will fine tune llms, create rag systems, transformer related tasks

What I do: I build NLP models and transformer-based systems from scratch in PyTorch not just wrappers around existing APIs. Services I offer: - Decoder-only / encoder-only transformers (custom architecture) - Fine-tuning open source LLMs (Llama, Mistral, GPT-2) - RAG pipelines (retrieval + generation) - Text classification, NER, summarization - On-device NLP (Android, quantized models) My background: - Completed Stanford CS224N - Built a GPT-style chess language model from scratch in PyTorch (custom tokenizer, multi-head attention, causal masking, positional embeddings) - NLP + Android development background rare combo for on-device ML You'll receive: - Clean, documented PyTorch code - Training + evaluation results - GitHub repo (on request) - Brief writeup of architectural decisions Message me before ordering if you're unsure which package fits your project.