AI Engineer

About the RoleYou will be designing, developing and deploying scalable generative AI solutions while managing and maintaining LLM deployments for internal AI tools, ensuring high availability and efficiency for internal teams, and training and deploying deep learning and LLMs models to production.ResponsibilitiesDesign and implement robust RAG pipelines to integrate Large Language Models (LLMs) with proprietary data sources, ensuring high accuracy and low latency in responses.Architect and maintain complex Apache Airflow DAGs to automate data ingestion, cleaning, embedding generation, and model retraining workflows.Manage and optimize Vector Databases (e.g., Pinecone, Milvus, Weaviate) for efficient storage and retrieval of high-dimensional embeddings.Maintain and monitor internal LLM deployments (hosting, scaling, and versioning), ensuring 99% uptime, managing GPU resources and optimizing inference for internal AI usage.Train, fine-tune, and deploy deep learning and LLM models to production environments, managing the full MLOps lifecycle from experimentation to serving.QualificationsAt least Bachelor's Degree in Computer Science, Software Engineering, Data Science, AI or related field8+ years in Software Engineering overallSolid experience in building LLM applications and RAG pipelines, in productionProduction experience with Python and DBs.Optimizing GPU compute resources (vLLM, TensorRT).Open-source contributions to AI/MLOps libraries.Advanced model fine-tuning (LoRA, PEFT).Good foundation in linear algebra and statistics.Required SkillsProgramming: Python (Must), Java (nice to have: GO, Scala)AI/ML: PyTorch, TensorFlow, Hugging Face.GenAI and LLMs: RAG, LangChain, LlamaIndex, RAG pipeline design. Nice to have: PEFT/LoRA fine-tuning.Data Processing: Apache Airflow, DAGs. Nice to have: PySpark or Ray.Databases: Vector DBs (Pinecone/Milvus), SQL/NoSQL, embedding models and semantic searchMLOps: Docker, Kubernetes, vLLM, Cloud (AWS/Azure/GCP), CI/CD pipelines. Nice to have: GPU resource management and optimization (CUDA).