Senior AI Research Engineer - Generative AI

About Ailytics

Ailytics develops AI-driven solutions aimed at creating safer environments by leveraging computer vision and predictive analytics. Their platforms are deployed globally, covering over 500 million square meters, helping organizations identify risks proactively, optimize processes, and save lives.

Role Overview

We seek a Senior AI Research Engineer to lead and expand our real-time generative AI capabilities currently in production. This role encompasses fine-tuning and adapting large language models (LLMs), vision-language models (VLMs), and extensive vision models, building new features, and establishing the foundational architecture. You will also manage deployment, taking models from experimentation to production without handoff.

Key Responsibilities

• Fine-tune, adapt, and experiment with foundation models (LLMs, VLMs, and vision models) to meet specific product needs and accuracy standards.
• Develop new generative AI functionalities from concept to deployment for customer usage.
• Enhance existing models focusing on accuracy, robustness, latency, and inference cost, ensuring real-time performance constraints are met.
• Lead the design and deployment architecture of generative AI systems integrated with the computer vision stack.
• Create rigorous evaluation frameworks and benchmarks to accurately assess model improvements.
• Curate and improve datasets through training and evaluation data refinement, augmentation techniques, and quality labelling, including handling edge cases.
• Define quality standards for generative AI functions and build a capable engineering team to scale these efforts.

Required Qualifications and Expertise

• More than 5 years in computer vision or machine learning engineering, including at least 2 years focusing on generative AI.
• Advanced proficiency in Python with deep understanding of data structures, algorithms, and software engineering best practices.
• Hands-on, up-to-date experience fine-tuning open-weight foundation models such as Qwen-VL, InternVL, LLaVA, Qwen, Llama, Mistral, DeepSeek, DINOv2, SAM, and Grounding DINO, with the ability to discuss challenges encountered.
• Comprehensive knowledge of LLM inference internals including attention mechanisms, KV-cache, and decoding strategies.
• Expertise in tokenization, managing long-context windows, and balancing context length against performance metrics like latency and throughput.
• Experience in deploying production models with multi-GPU sharding, memory and throughput optimization using tools like vLLM, SGLang, or TensorRT-LLM, with the ability to troubleshoot bottlenecks effectively.
• Preferred experience running inference on edge or on-premise hardware (e.g., NVIDIA Jetson), including quantization and real-time optimization under constrained computational resources.
• Track record of building or shipping scalable LLM-based applications, along with strong familiarity with PyTorch and the Hugging Face ecosystem, including expertise in LoRA/QLoRA, supervised fine-tuning, quantization, and distillation.
• Good understanding of multi-agent systems and agentic AI including tool-use, orchestration, and multi-step agent design.
• Strong skills in evaluation and experiment design to obtain reliable model assessments.
• Experience with multimodal systems combining vision and language techniques.
• Degree in Computer Science, Electrical Engineering, or a related discipline (Bachelor's, Master's, or PhD).
• Proven architectural judgment and experience owning technical design decisions.
• Leadership aptitude with experience mentoring engineers and enthusiasm for growing a team and function beyond individual contribution.

Why Join Us?

• Make a meaningful impact at the convergence of AI and safety, powering enterprise tools used daily in critical environments across Asia and beyond.
• Opportunity to shape both product direction and AI capabilities within an early-stage startup environment.
• Ownership of the generative AI function including architecture, roadmap, and scaling.
• Support for continuous learning, including a budget for education and encouragement to participate in leading conferences and publications.
• Culture valuing high standards, transparency, and collaborative low-ego teamwork.

Additional Information

This role is based in Singapore and is an onsite full-time position. You will have direct collaboration with the CTO, product managers, and senior leadership. Candidates can expect to work with uniquely large and proprietary industrial video datasets unavailable publicly, providing distinctive technical challenges and opportunities.