My current Natural Language Processing pipeline is delivering decent results, but I know it can do far better in both accuracy and runtime. I’ve already gathered the training data, built an initial model, and wired it into the product. Now I need a specialist who can step in, audit the existing code, spot bottlenecks, recommend smarter architectures, and fine-tune the model so it performs reliably at scale. The stack in play includes Python, PyTorch, Hugging Face transformers, and a small dose of TensorFlow for legacy components. You’ll have access to clean, labeled datasets, a GPU-ready environment, and evaluation scripts. What I’m counting on is your deep experience distilling larger language models, optimizing inference, and tightening up preprocessing so latency stays low without sacrificing accuracy. Please highlight similar NLP projects you’ve tackled, the concrete gains you achieved (e.g., accuracy lift, speedup, memory reduction), and any tooling tricks you relied on. That real-world experience will be the key factor in moving forward. Deliverables • A refactored or newly trained model that demonstrably outperforms the baseline on my validation set • A brief technical report summarizing changes, metrics, and next-step suggestions • Updated deployment scripts or Dockerfile so the upgrade slots straight into production Acceptance will be based on hitting the target metrics we’ll define together (precision, recall, and 2× faster inference as an initial goal) and clean, reproducible code.