AI Architect (GenAI, Agentic AI, on-premise)

Role: AI Architect

Location: Bengaluru/ Hyderabad/ Chennai/ Pune/ Mumbai/ Gurugram (Hydrid)

Job Type: Permanent

Technical Requirements:

• Agentic AI: Proficiency in Python or Java.

• Infrastructure: Direct experience with on-premise GPU-as-a-Service.

Core Infrastructure & Deployment

• Containerization: Docker for packaging models.

• Orchestration: Kubernetes (K8s) for cluster management.

• GPU Architecture: Knowledge of NVIDIA CUDA, ROCm, and vGPU allocation.

• Model Serving: Deployment via vLLM, Ollama, TGI, or Triton Inference Server.

Model Management & Optimization

• Quantization: AWQ, GPTQ, or GGUF for local VRAM optimization.

• Fine-Tuning: PEFT and LoRA for local domain adaptation.

• Open-Source Selection: Experience with Hugging Face models (e.g., Llama, Mistral, Phi).

• Hardware Benchmarking: Memory bandwidth and compute requirement calculations.

Agent Architecture & Integration

• Framework Mastery: LangChain, LangGraph, or AutoGen.

• Vector Databases: Pinecone, Qdrant, Milvus, or PGVector.

• RAG Pipelines: Building robust systems without cloud APIs.

• State Management: Handling complex agent memory loops locally.

On-Premise Security & Networking

• Air-Gapped Workflows: Dependency and model installation without internet access.

• Data Privacy: Local data governance and access controls.

• Local Auditing: Private logging and guardrails (e.g., NeMo Guardrails).

Software Engineering & DevOps

• Languages: Python, Rust, C++, and Java Microservices.

• CI/CD Pipelines: GitLab CI or Jenkins.

• Monitoring: OpenTEL, Prometheus, and Grafana.