Startups training their own language models and working on reasoning

Abacus.AI is a startup that specializes in developing and training its own language models, as evidenced by their offerings such as Custom Fine-Tunes, LLM and Retrieval APIs, and the development of custom LLMs tailored to specific business needs. Additionally, they are working on reasoning capabilities within their technology, as indicated by their AI agents that can automate complex tasks and workflows, and their use of generative AI and other novel neural net techniques to build predictive applied AI systems.

NuMind is a startup (founded in June 2022, team size of 6, and part of Y Combinator S22 batch). The company is developing and training its own language models, as evidenced by their work on foundation models for entity recognition and task-specific foundation models with GPT-4. Additionally, they are working on reasoning capabilities within their technology, as indicated by their focus on NLP tasks such as classification, sentiment analysis, and entity recognition, which require reasoning capabilities.

JUST ADD AI is a startup that develops custom language models and offers solutions in generative AI, including reasoning capabilities such as predictive maintenance, anomaly detection, and intelligent alarm filtering.

Mendel.ai is a startup that has developed its own proprietary clinical AI, which includes training machine learning models on real medical records. The company also works on clinical reasoning capabilities within their technology.

Metabob is a startup that utilizes proprietary graph neural networks (GNNs) and large language models (LLMs) to detect, explain, and fix coding problems. This indicates that they are developing and training their own language models. Additionally, their technology works on reasoning capabilities as it generates context-sensitive problem explanations and resolutions, and detects complex problems across codebases.

Plasticity is a startup (Y-Combinator S17 batch) that develops its own language models, as evidenced by their Sapien Language Engine, Cortex Knowledge Graph, and Lingua Dialogue Engine. These models support complex reasoning capabilities such as inference questions, intent parsing, and multi-turn conversations with context management.

OSlash is a startup founded in 2020 and has raised $7.5 million in funding. The company is developing its own language models, as evidenced by its use of the latest embedding model based on the GPT-3 tokenizer and various search-specific algorithms. Additionally, OSlash is working on reasoning capabilities within their technology, as indicated by features like the TaskSynth classifier model that generates workflows from natural language commands and the IntelliDetect model that classifies user intent with high accuracy.

Neuron7 is a startup that has raised $10M in Series A funding. The company uses large language models (LLMs) trained to understand service use cases and specific industry terminology, which indicates they are developing or training their own language models. Additionally, their solutions provide guided step-by-step resolutions for complex issues and capture and learn from every interaction to continually improve resolution accuracy, suggesting they are working on reasoning capabilities within their technology.

Suki is a startup that develops and trains its own language models, as evidenced by its use of large language models (LLMs) and generative AI. Additionally, Suki's AI assistant works on reasoning capabilities, such as answering questions and recommending codes.