AI Engineer - NLP/Generative AI in Marathi, Maharashtra, भारत

Job Description : Role : AI EngineerCompany : Webnyay Private LimitedLocation : PuneExperience : 5+ years, with hands-on NLP and Generative AI workEducation : B.E./B.Tech/M.Tech with a specialisation in AI/MLCore focus : Classification, NLP and GenAI workflows, all self-hostedEmployment type : Full-time On-SiteBYOD : Bring your own deviceWorking Hours : 9 : 30am-7 : 30pm (Mon-Fri, Saturday is half day and work from home)Required qualifications and skills : - B.E./B.Tech/M.Tech with a specialisation in AI/ML, or equivalent.- 5+ years in ML engineering, including enterprise classification, NLP and Generative AI workflows taken to production.- Strong Python, with PyTorch and Hugging Face Transformers.- Hands-on fine-tuning of open-weight LLMs (for example with LoRA or QLoRA) and building retrieval-augmented generation (RAG) pipelines.- Experience self-hosting models with an inference server such as vLLM, TGI or Ollama, on GPU or CPU.- Solid grounding in text classification, named entity recognition and summarisation, including evaluation metrics and calibration.- Experience exposing models as APIs (FastAPI or similar) and packaging them with Docker.- Understanding of PII handling and data protection in ML https://jobeax.com/link/H0cIAD3DywuYVRu3 to have : - NLP for Indian languages (Hindi and regional languages), including speech-to-text.- Experience in BFSI or legal text, team lead.- MLOps tooling for experiment tracking, model registry and drift monitoring (for example MLflow).- Experience with vector databases such as pgvector, Qdrant or Milvus.- Exposure to the DPDP Act 2023, RBI guidance on AI use, or model risk management.- Published work, open-source contributions or Kaggle results in https://jobeax.com/link/pCHZgfj4ucIsKPof we offer : - Ownership of a GenAI system in production at a public sector undertaking.- A small, senior team where your models ship to production quickly.- Work at the intersection of GenAI, regulation and consumer protection. (ref:hirist.tech)