Senior LLM / Multimodal AI Engineer

Ncs3; Singapore; On-site; SGD 180,000 - 260,000; Full time

Job summary

NCS invites an experienced Senior LLM / Multimodal AI Engineer to shape next-generation foundation models and multimodal systems at scale in Singapore. You will architect, train, evaluate, and deploy VLMs and LLMs processing billions of UGC and ad videos across advertising ecosystems.

Responsibilities include data pipelines, supervised fine-tuning, PEFT/LoRA techniques, and scalable retrieval systems. You will collaborate with AI Solution Architects and Data Engineers to deliver impactful AI

Qualifications

• 4+ years hands-on experience in ML systems development (NLP/CV/LLMs/Multimodal).
• Proven track record from research to production deployment.

Responsibilities

• Data curation and multimodal pre-training.
• Model training, fine-tuning, and deployment for VLMs/LLMs.
• Design and implement retrieval-augmented generation (RAG) pipelines.
• Optimize inference performance (vLLM, TensorRT-LLM, Triton).
• Evaluate models with online/offline benchmarks and monitoring.

Skills

Deep learning

Transformers

Multimodal fusion

PyTorch

Hugging Face

DeepSpeed

PEFT

LlamaFactory

Python

SQL

Spark

Linux

Git

FAISS

vLLM

TensorRT-LLM

Triton

Tools

FAISS

TensorRT-LLM

Triton

Langfuse

Weights & Biases

MLflow

Job description

Join NCS as a Senior LLM / Multimodal AI Engineer and spearhead the development of next-generation foundation models and multimodal understanding systems. With 4+ years of hands-on experience, you will architect, train, evaluate, and deploy large-scale Vision-Language Models (VLMs) and Large Language Models (LLMs) that process billions of user-generated content (UGC) and advertising videos. You will transform cutting-edge research into industrial-scale recommendation and advertising systems that deliver measurable business impact.

What will you do?

Data Curation & Multimodal Pre-Training

• Architect and scale automated data pipelines that integrate video frames, images, OCR, ASR, and text for VLM pre-training and fine-tuning
• Drive LLM-assisted annotation workflows and synthetic data generation strategies to minimise manual labeling overhead
• Engineer robust data quality frameworks ensuring representative, accurate, and complete training datasets

Model Training & Fine-Tuning

• Reproduce and optimise foundation model series (e.g., LLaVA, Qwen-VL family) for production environments
• Execute supervised fine-tuning (SFT) and post-training for domain-specific content ecosystems including gaming, movies, automotive, and commercial advertising
• Implement advanced training techniques including PEFT, LoRA, and quantization-aware training

Multimodal RAG & Knowledge Bases

• Design and deploy hierarchical Retrieval-Augmented Generation (RAG) pipelines that enrich video understanding and enable in-context retrieval
• Build scalable vector search indexes (e.g., FAISS) supporting billions of embeddings
• Optimise retrieval strategies including hybrid search, re-ranking, and query expansion

Inference Optimization & Deployment

• Optimise large-scale VLM and LLM serving using vLLM, TensorRT-LLM, Triton, and ONNX Runtime
• Implement dynamic batching, feature caching, and asynchronous scheduling to maximise GPU throughput
• Monitor and tune production inference systems for latency, cost, and quality trade-offs

Evaluation & Agent Workflows

• Establish comprehensive online/offline evaluation benchmarks measuring generative quality, aesthetic completeness, and vertical task precision/recall
• Contribute to intelligent tag platform agents and agentic workflow orchestration
• Conduct continuous LLM-as-Judge evaluation on production traces

An Ideal Candidate Should Possess:

Experience

• 4+ years of hands-on experience in machine learning system development, with demonstrated focus on NLP, Computer Vision, LLMs, or Multimodal AI
• Proven track record of independently driving end-to-end ML initiatives from research exploration to high-throughput production deployment

Core Technical Skills

• Deep expertise in deep learning fundamentals, transformer architectures, and multimodal fusion techniques
• Proficiency in deep learning frameworks: PyTorch, Hugging Face ecosystem, DeepSpeed, PEFT, LlamaFactory
• Strong programming capabilities in Python, SQL, Spark, and Linux/Git workflows
• Hands‑on experience with vector search and retrieval infrastructure (FAISS, Multimodal RAG)
• Familiarity with LLM/VLM serving and quantization toolkits (vLLM, TensorRT-LLM, Triton)

Preferred Qualifications

• Industrial experience in large-scale recommender systems, advertising technology, or content moderation/understanding ecosystems
• Hands‑on expertise with reinforcement learning from human feedback (RLHF) or preference optimization (e.g., DPO, GRPO)
• Published research or competitive achievements in top AI/ML venues or challenges (e.g., CVPR, NeurIPS, ICLR, Kaggle)
• Experience with cloud ML platforms (AWS SageMaker, Azure ML, Google Vertex AI)
• Knowledge of AI observability tools (Langfuse, Weights & Biases, MLflow)

Collaborate with NCS

At NCS, you'll work alongside cross-functional teams including AI Solution Architects, Data Engineers, and Product specialists to deliver transformative multimodal AI solutions. Your expertise will directly enhance our Video Intelligence Platform and support AI innovation across Public Service, Defence and Homeland Security, Healthcare and Transport, and other key sectors.

Join us in building a better future where your foundation model expertise transforms how organisations understand and leverage video content at scale.

We are driven by our AEIOU beliefs-Adventure, Excellence, Integrity, Ownership, and Unity-and we seek individuals who embody these values in both their professional and personal lives. We are committed to our Impact: Valuing our clients, Growing our people, and Creating our future.

Together, we make the extraordinary happen.

Learn more about us at ncs.co and visit our LinkedIn career site.