Algorithm Researcher — Agentic Vision

Origen Tech; Singapore; On-site; SGD 120,000 - 190,000; Full time

Job summary

Origen Tech seeks a research scientist to lead post-training R&D for vision-language models. You will own end-to-end algorithm development for post-training, exploring SFT and RLHF, PPO, GRPO, and expanding the capabilities of vision foundation models.

You will advance visual perception and multimodal reasoning, reduce model hallucinations with detection pipelines and data cleaning, and improve real-world user interactions with robust evaluation in multi-turn image-text scenarios.

Qualifications

• Master's degree or higher in Computer Science, Software Engineering, AI, or related field.
• Strong understanding of Transformer architectures and LLM training principles.
• Experience in post-training domains including LLM Alignment, RLHF, and memory/RAG.
• Experience with large-scale model training and inference.

Responsibilities

• Own the full-stack algorithm R&D for post-training of vision-language models.
• Explore SFT and reinforcement learning alignment methods (RLHF, PPO, GRPO).
• Improve low-level visual perception and spatial understanding.
• Reduce hallucinations and build robust hallucination detection pipelines.
• Enhance multi-turn image-text interaction and real-world user scenarios.

Skills

Transformer

LLM Training

RLHF

Reward Modeling

Memory/RAG

Python

PyTorch

TensorFlow

Megatron-LM

DeepSpeed

vLLM

Publications

HuggingFace

Teamwork

Education

Master's degree or higher in CS/AI

Tools

Megatron-LM

DeepSpeed

vLLM

HuggingFace

Job description

Job Description

• Lead Post-Training R&D System: Take ownership of the full-stack algorithm R&D for post-training of Vision-Language Models (VLMs). Deeply explore Supervised Fine-Tuning (SFT) and reinforcement learning alignment algorithms (e.g., RLHF, PPO, GRPO),continuously pushing the boundaries and elevating the capability ceiling of vision foundation models.


• Advance Visual Foundation Perception: Dedicate to optimizing the model's low-level visual perception, fine-grained understanding, and spatial awareness. Proactively explore Visual Agent solutions including \"thinking with images,\" endowing models with deeper comprehension of and execution capability in the physical world.


• Enhance Large Model Reliability: Strive to substantially reduce hallucinations in multimodal large models. Investigate the root causes of hallucination, build comprehensive hallucination detection pipelines, implement refined data cleaning strategies, and spearhead the R&D of reinforcement learning-based hallucination removal solutions.


• Empower User Interaction Scenarios: Align and optimize models for real-world user scenarios, strengthening the model's logical reasoning, instruction following, and detail perception capabilities in multi-turn image-text interactions. Build an online badcase monitoring and data flywheel feedback loop to continuously refine and deliver exceptional interaction experiences.

Skills Required

• Master's degree or above in Computer Science, Software Engineering, Artificial Intelligence, or related fields.


• Deep understanding of the Transformer architecture and large language model training principles. Extensive research and hands-on experience in at least one post-training domain, including LLM Alignment, RLHF, Reward Modeling, Personalized LLM, and Memory/RAG mechanisms.


• Rich practical experience in user profiling, recommendation systems integrated with large models, or long-context/long-term memory model optimization.


• Solid algorithm foundation and engineering implementation skills, proficient inPython, and familiar with deep learning frameworks such as PyTorch andTensorFlow.


• Hands-on experience with distributed training, familiar with large-scale training and inference frameworks including Megatron-LM, DeepSpeed, and vLLM.


• Experience in training or fine-tuning models at the tens-of-billions to hundreds-of-billions parameter scale.


• Excellent research acumen, with publications at top-tier conferences (NeurIPS, ICLR, ICML, ACL, EMNLP, etc.) or significant contributions to high-impact open-source projects (e.g., on HuggingFace).


• Strong technical passion and self-drive, adept at analyzing and solving complex problems, with good teamwork and communication skills.