Position : Data scientist or Machine Learning with Research (4–6 years of hands-on experience in research )
Location : San Jose, California
Job type : contract
H1b,GC,USC
Must Need master’s degree
They would also prefer candidates with a Master's or PhD in a relevant field.
who works below the framework layer - custom loss functions, DPO/KTO and other post-training methods, training-loop and optimizer-level work, and distributed training systems.
Client’s Machine Learning hiring requirements and identify a suitable candidate profile to support the development of advanced AI training, post-training, and model optimization infrastructure.
Client Background & Current Initiative
Client shared an overview of his experience and current AI initiatives:
• Over 25 years of experience in Machine Learning and AI.
• Has spent the last 10 years building investigation and consulting systems across multiple industries.
• Has accumulated a large volume of proprietary, machine-learning-ready data and is converting it into a comprehensive post-training platform.
• Plans to commercialize and license this data to major AI labs and organizations.
• Operates in a highly secure environment with sensitive customer data and therefore runs a fully offline AI infrastructure.
• Has built and maintains his own NVIDIA Grace Blackwell-based GPU cluster with high-speed networking infrastructure.
Has developed multiple internal AI systems, including Custom evaluation platform
Knowledge graph pipelines
GraphRAG implementations
Named Entity Recognition (NER) models
Multi-model orchestration framework
AI assistant ensemble systems
Synthetic data generation workflows
Custom model evaluation and monitoring systems
Client explained that his long-term vision is to build a recursive AI platform that continuously learns from analyst interactions and real-world usage data.
Hiring Requirement
Primary Requirement
Client is looking for a highly experienced Machine Learning Research Engineer who can assist with:
• Custom loss function development
• Model training and post-training pipelines
• Foundation model adaptation and fine-tuning
• Mixture of Experts (MoE) architectures
• Model weight optimization
• DPO (Direct Preference Optimization)
• ATO and other post-training methodologies
• Synthetic data generation
• Model evaluation and benchmarking
• Training infrastructure development
• Experimentation with adapters and specialized model architectures
Desired Candidate Profile
The ideal candidate should:
• Have a deep understanding of model internals.
• Be comfortable working directly with model weights, training mechanisms, and optimization techniques.
• Possess strong hands-on implementation experience.
• Understand foundation models and modern post-training approaches.
• Have a research-oriented mindset and be comfortable experimenting with new techniques.
• Ideally have exposure to academic research and publications, though practical execution is the primary requirement.
Client emphasized that he is looking for someone who can work “under the hood” rather than someone who only utilizes existing AI tools and frameworks.
Research & Publication Plans
• Client intends to publish research papers in the future based on the work being conducted.
• Research publication is currently secondary to product development and business growth.
• He already has access to academic and research communities that can support publication efforts.
• Candidate participation in future research publications may be beneficial but is not a mandatory requirement.
• The immediate priority is finding someone who can contribute directly to model development and optimization.
Client highlighted the importance of reviewing practical work before moving forward with candidates.
Key evaluation criteria include:
• GitHub repositories and publicly available projects.
• Evidence of hands-on machine learning and research work.
• Technical depth in model training and optimization.
• Experience working on real-world AI systems.