Custom English LLM Development
Budget: ₹1,500 – ₹12,500 INR
I plan to create a domain-specific Large Language Model trained solely on English text. You will manage the full pipeline—from raw-text ingestion and cleaning through model architecture selection, distributed training and final evaluation. Open-source frameworks such as PyTorch, TensorFlow or the Hugging Face stack are welcome as long as the end result is reproducible and can be deployed behind a simple REST or gRPC API.
Key deliverables
• Curated and documented English-text dataset (with clear licensing)
• Training scripts and configuration files ready for re-run on our GPU cluster
• Trained model weights and tokenizer
• Evaluation report covering perplexity, BLEU/ROUGE and safety checks
• Lightweight inference wrapper with example requests and responses
Acceptance criteria
• All code runs end-to-end in a fresh environment with provided instructions
• Model meets or exceeds agreed test metrics on a held-out set
• No personally identifiable or restricted data appears in the final artifacts
When you submit your proposal, showcase past work that demonstrates your experience with LLMs or comparable NLP systems; concise links or repos are ideal. I will shortlist based on proven delivery of similar projects and clarity of technical approach.
Key deliverables
• Curated and documented English-text dataset (with clear licensing)
• Training scripts and configuration files ready for re-run on our GPU cluster
• Trained model weights and tokenizer
• Evaluation report covering perplexity, BLEU/ROUGE and safety checks
• Lightweight inference wrapper with example requests and responses
Acceptance criteria
• All code runs end-to-end in a fresh environment with provided instructions
• Model meets or exceeds agreed test metrics on a held-out set
• No personally identifiable or restricted data appears in the final artifacts
When you submit your proposal, showcase past work that demonstrates your experience with LLMs or comparable NLP systems; concise links or repos are ideal. I will shortlist based on proven delivery of similar projects and clarity of technical approach.