Model Expertise : Work with transformer models (GPT, BERT, T5, RoBERTa, etc.) across NLP tasks including text generation, summarization, classification, and translation.
Model Fine-Tuning : Fine-tune pre-trained models on domain-specific datasets to optimize for summarization, text generation, question answering, and related tasks.
Prompt Engineering : Design, test, and iterate on contextually relevant prompts to guide model outputs for desired performance.
Instruction-Based Prompting : Implement and refine instruction-based prompting strategies to achieve contextually accurate results.
Learning Approaches : Apply zero-shot, few-shot, and many-shot learning methods to maximize model performance without extensive retraining.
Reasoning Enhancement : Leverage Chain-of-Thought (CoT) prompting for structured, step-by-step reasoning in complex tasks.
Model Evaluation : Evaluate model performance using BLEU, ROUGE, and other relevant metrics; identify opportunities for improvement.
Deployment : Deploy trained and fine-tuned models into production environments, integrating with real-time systems and pipelines.
Bias & Reliability : Identify, monitor, and mitigate issues related to bias, hallucinations, and knowledge cutoffs in LLMs.
Collaboration : Work closely with cross-functional teams (data scientists, engineers, product managers) to design scalable and efficient NLP-driven solutions.
Must-Have Skills :
7+ years of overall experience in software / AI development with at least 2+ years in transformer-based NLP models .
4+ years of hands-on expertise with transformer architectures (GPT, BERT, T5, RoBERTa, etc.).
Strong understanding of attention mechanisms, self-attention layers, tokenization, embeddings, and context windows .
Proven experience in fine-tuning pre-trained models for NLP tasks (summarization, classification, text generation, translation, Q&A).
Expertise in prompt engineering , including zero-shot, few-shot, many-shot learning, and prompt template creation.
Experience with instruction-based prompting and Chain-of-Thought prompting for reasoning tasks.
Proficiency in Python and NLP libraries / frameworks such as Hugging Face Transformers, SpaCy, NLTK, PyTorch, TensorFlow .
Strong knowledge of model evaluation metrics (BLEU, ROUGE, perplexity, etc.).
Experience in deploying models into production environments .
Awareness of bias, hallucinations, and limitations in LLM outputs .
Good to Have :
Experience with LLM observability tools and monitoring pipelines.
Exposure to cloud platforms (AWS, GCP, Azure) for scalable model deployment.
Knowledge of MLOps practices for model lifecycle management.