Soul AI builds reinforcement learning from human feedback (RLHF) and custom AI solutions for enterprise clients. The stack—Python, Go, Kubernetes, RAG, plus Claude and Cursor—reflects a lean, infrastructure-forward team focused on model training and evaluation. Active projects span LLM training, video model refinement, code annotation, and financial AI, while pain points cluster around model accuracy, hallucination detection, and safety in generated code—suggesting deep work on production AI reliability rather than experimentation.
Tech Stack 106 technologies
Core StackGo Docker Kubernetes Airtable RAG Python Apache Spark Hadoop AWS Laravel PHP Cursor JavaScript React Vue Angular Webpack Redux Jest Rust Excel Notion SQL GCP Azure Symfony PHPUnit Claude HTML CSS+74 more
AdoptingReact
What Soul AI Is Building
◆Challenges
- Improving ai model accuracy
- Improving accuracy of ai-generated frontend solutions
- Legal annotation quality
- Evaluating ai-generated engineering solutions
- Improving ai-generated rust code quality
- Reducing bias in ai-generated investment analyses
- Improving accuracy & performance of sota models
- Detecting hallucinations in research
- Ai-generated embedded code quality improvement
- Safety of ai-generated embedded solutions
▲Active Projects
- Llm training projects
- Training and evaluating advanced generative video models
- Ai-generated php code annotation
- Building high-quality embedded systems datasets
- Building ai foundational models for engineering and infrastructure tasks
- Ai prompt annotation
- Ai evaluation of m&a financial modeling
- Advanced ai evaluation initiatives for equity research
- Ai model refinement
- Financial rag system development
Hiring Activity
Minimal110 roles · 0 in 30d
Department
Seniority
Company intelligence
