<p><p><b>Role Overview :</b><br/><br/>You will lead the fine-tuning and evaluation of open-source LLMs (e.g., Llama, Mistral, Qwen).
This role requires hands-on skills in Python, PyTorch, Hugging Face, and distributed training :</b></p><p><br/></p>- Fine-tune and adapt open-source LLMs using LoRA/QLoRA.<br/><br/></p><p>- Build pipelines for supervised fine-tuning and preference optimization (DPO/RLHF).<br/><br/></p><p>- Evaluate model performance using standard and domain-specific benchmarks.<br/><br/></p><p>- Optimize inference with quantization and serving frameworks (vLLM, TGI).<br/><br/></p><p>- Collaborate with data engineers on preprocessing and :</b></p><p><br/></p>- Experience: 3-6 years in ML/NLP<br/><br/></p><p>- Strong coding in Python and PyTorch.<br/><br/></p><p>- Experience with Hugging Face, DeepSpeed, or similar frameworks.<br/><br/></p><p>- Familiarity with distributed training and GPU optimization.<br/><br/></p><p>- Knowledge of transformers, attention mechanisms, and NLP workflows.<br/><br/></p><p>- Bonus: Experience in domain-specific fine-tuning (finance, legal, etc.).</p><br/></p> (ref:hirist.tech)