Data Scientist

About the Role

We are looking for an AI Engineer to deploy and build AI applications on our in-house GPU servers in a fully air-gapped environment. You will work with LLMs, vision-language models, Stable Diffusion, and audio models, and build API services on top of them. You will work under the guidance of the senior team and own well-defined modules end to end.

Key Responsibilities

• Deploy and serve AI models locally on GPU servers with no internet access.
• Build AI applications similar to ChatGPT and Claude using LLMs and VLMs (chat assistants, document Q&A, multimodal apps).
• Maintain architecture documents, runbooks, and best practices.
• Monitor GPU utilization, model performance, and service health; lead production troubleshooting.
• Inference and test models such as Mistral, Llama, and GPT-class models.
• Deploy and run LLM, VLM & Stable Diffusion (image generation) and audio models (speech-to-text, text-to-speech).
• Develop and maintain APIs using frameworks like FastAPI to expose models to internal users and applications.
• Build RAG pipelines using embeddings models and vector databases.
• Support inference optimization: quantization, batching, and GPU memory management.
• Set up offline packages, dependencies, and model weights in the air-gapped environment.
• Write clear documentation for code, APIs, and deployment steps.
• Participate in design discussions and code reviews.

Requirements & Tech Stack

• Programming: Strong Python.
• Model Serving: vLLM, SGLang, TensorRT-LLM.
• Vector databases: Qdrant, FAISS, and Milvus.
• API development: FastAPI (async, streaming responses, validation, authentication basics).
• LLM and VLM: Hands-on experience with Mistral, Llama, or similar models.
• Model serving: Experience with at least one of vLLM, SGLang, TensorRT-LLM.
• Generative AI: Working knowledge of LLM, VLM & Stable Diffusion (Diffusers/ComfyUI) and audio models (Whisper, TTS).
• Frameworks: PyTorch, Hugging Face Transformers.
• GPU basics: CUDA, NVIDIA drivers, running models on single and multi-GPU setups.
• DevOps basics: Linux, Docker, Git.
• Air-gapped awareness: Comfortable working without internet and managing offline dependencies.

Good to Have

• LangChain or LlamaIndex.
• Vector databases: Qdrant, Milvus, Chroma.
• Fine-tuning experience (LoRA/QLoRA).
• Quantization formats (GGUF, AWQ, GPTQ).
• Gradio or Streamlit for quick UIs.
• Prometheus and Grafana.

Qualifications

• Bachelor's degree in Computer Science, AI/ML, Data Science, or a related field.
• Experience taking at least one AI project to production or a working deployment.

What We Offer

• Hands-on work with high-end GPU infrastructure.
• Exposure to the latest open-source and generative AI models.
• Clear growth path to Senior AI Engineer.