Hyper Lychee Labs
Website:
hyperlycheelabs.com
Job details:
ENGAGEMENT: Full-time
WORK MODE: India (Remote / Off-site – quarterly on-site visits to client site in Saudi Arabia)
JOB DESCRIPTION:
Key Responsibilities
- Own the operationalisation of both traditional ML models and generative AI / LLM systems in a fully on-premise environment.
- Deploy, monitor, and maintain LLM and agentic AI infrastructure running on client-owned GPUs.
- Manage traditional MLOps responsibilities (model deployment, versioning, monitoring) alongside newer LLMOps workflows – this is a blended role, not LLM-only.
- Work closely with the AI Infrastructure Architect to implement designed architectures for LLM/GPU management, parallelization, and agent orchestration.
- Manage GPU resource allocation and token/context handling for on-premise LLM deployments (distinct from cloud-based token economics).
- Support integration of MLOps pipelines with agentic and generative AI use cases as they move from design to production.
Must-Have Qualifications
- Practical experience with LLM/generative AI deployment and management
- Working knowledge of on-premise infrastructure management – this is not a cloud-native role. Candidates should understand how GPUs are provisioned, shared, and monitored outside of AWS/Azure/GCP-managed environments.
- Understanding of how LLMs are stored and served across GPUs, and of parallelization approaches for splitting models/workloads across GPU resources.
- Comfort working as part of a broader AI infrastructure team that includes an architect and, eventually, agent-orchestration workflows.
- Secondary qualifications: Support MLOPs teams when necessary. Hands-on experience operationalizing ML models in production (traditional MLOps).
Nice-to-Have
- Exposure to banking, financial services, or other highly regulated environments.
- Familiarity with agentic AI frameworks and orchestration patterns.
- On-premise GPU cluster management experience at scale.
Skills Required
- Python, TensorFlow/PyTorch, GPU management (CUDA, NVIDIA tooling).
- CI/CD for ML, Docker/Kubernetes, model-serving frameworks (Triton, TorchServe, vLLM).
Preferred Skills
- Agentic AI frameworks (LangChain, LlamaIndex).
- GenAI deployment (quantization, fine-tuning, RAG).
- On-prem orchestration (Slurm/Kubeflow/Ray).
Working Arrangement
This is a technical, execution-focused role with limited stakeholder-facing requirements, which is why it is structured as off-site with periodic on-site visits (approximately once per quarter) rather than a full-time on-site posting.
ABOUT US
Hyper Lychee Labs is an IT services and consulting company committed to delivering innovative and impactful solutions to a diverse client base, including Fortune 500 companies, the Government of India, and forward-thinking startups. With a focus on leveraging technology for transformation, the company provides custom software development, complex systems implementation, and strategic consulting. Our team of talented professionals collaborates to ensure exceptional results tailored to clients’ needs. Dedicated to fostering a culture of creativity, diversity, and inclusion, Hyper Lychee Labs strives to empower businesses to thrive in a rapidly evolving world.
Click on Apply to know more.