Key Responsibilities : - Model Development & Fine-tuning : Lead the selection, architecture, and fine-tuning of LLMs for high-accuracy document classification and data extraction from unstructured sources.- End-to-End Extraction Pipelines : Design and optimize workflows to extract structured data from diverse documents such as GST certificates, Udyam registrations, FSSAI licenses, and property-related legal documents.- Performance Evaluation : Conduct rigorous end-to-end analysis and result metric evaluation (e.g., Precision, Recall, F1-score) to ensure model reliability across production batches.- OCR Integration : Collaborate with engineering teams to integrate central OCR services and refine extraction keys for specific business release cycles.- Production Deployment : Build and maintain high-performance REST APIs and microservices using Python (FastAPI/Flask) within AWS environments.- Advanced AI Tasks : Implement techniques for resolving name variations, identifying corporate hierarchies,
and normalizing unstructured text from SMS and bureau records.Required Skills & Qualifications : - Experience : 2 - 4 years of hands-on experience in AI/ML engineering, with a proven track record in document intelligence or specialized NLP.- LLM Expertise : Demonstrated experience fine-tuning LLMs (e.g., Llama, Mistral, BERT-based architectures) for classification and information extraction tasks.- Technical Stack : Solid proficiency in Python, including data handling libraries like Pandas and NumPy, and ML frameworks such as PyTorch or TensorFlow.- Production Knowledge : Familiarity with CI/CD pipelines, MLOps, Docker, Kubernetes, and cloud platforms like AWS (specifically Lambda or EC2-based API deployments).- Database Proficiency : Solid understanding of SQL (MySQL) and NoSQL databases for handling large-scale document metadata.- Education : BE/Btech in Computer Science from Tier1 colleges. (ref:hirist.tech)