# Hyde

> Hyde is the integrated training and inference platform that lets enterprises turn their proprietary data and expertise into specialist AI models that outperform the frontier. The thesis: the next era of enterprise AI is small, owned, and highly specialized, and the companies that own their intelligence will outperform the ones that rent it.

Hyde (Hyde Inc., also operating as Build with Hyde) partners with enterprises to extract deep institutional context from their operations and package it into post-trained sovereign AI models. The platform spans three integrated stages: preparing training-grade data and simulation environments, efficient post-training, and self-tuning inference that improves with each interaction. Models stay private, with no leakage of customer IP or data and no reliance on external model providers.

Reported outcomes from partners include up to 10x lower inference cost and latency at scale, up to 7x lower long-term cost of ownership, and up to 100x faster training efficiency via GPU-optimized pipelines. Capabilities span reasoning, forecasting, code, browser, voice, and vision. Hyde is backed by General Catalyst and Bain Capital Ventures, and partners with tier-1 enterprise AI firms like Palantir, Cognition, NVIDIA, Bain & Company, BCG and AWS.

Hyde was founded by Palantir's prior head of special projects and has a high talent density of AI talent from firms like Deepmind, Jane Street, and McKinsey.

## Core pages

- [Home](https://hyde.ai/): Company overview, value proposition, and partner results across reasoning, forecasting, code, voice, and vision.
- [Product](https://hyde.ai/product): The training and inference platform in detail, covering environment and data preparation, training techniques and trajectory engineering, and self-tuning inference with version control.
- [Blog](https://hyde.ai/blog): Research notes, case studies, and company updates.
- [Careers](https://hyde.ai/careers): Open roles and the three operating principles (radical ownership, idea meritocracy, rigor and velocity).

## Product details

- [Environment](https://hyde.ai/product): Context-specific benchmarks distilled from expert knowledge, a synthetic data engine to overcome data scarcity, and reusable modular assets (datasets, trajectories, evaluators).
- [Training](https://hyde.ai/product): GPU-optimized pipelines, support for advanced post-training techniques and real-world simulations, and real-time trajectory engineering to update policies on the fly.
- [Inference](https://hyde.ai/product): Cost, latency, and efficiency tuning, a tightly integrated inference-training loop for continuous improvement, and comprehensive version control with rollbacks.

## Case studies and research

- [Own your Intelligence](https://hyde.ai/blog/own-your-intelligence): Hyde's manifesto on why enterprises should own rather than rent their AI.
- [US stock exchange](https://hyde.ai/blog/us-stock-exchange): A national US stock exchange built a sub-10B-parameter specialist model for regulatory reasoning and reporting, approaching full accuracy on SEC reports with a $20M+ white-label revenue opportunity.
- [Tata Motors: 4x Sales Conversion](https://hyde.ai/blog/tata-motors-4x-sales-conversion): The $50B automaker built specialist models to automate car sales, with a 4x improvement in sales conversion and a leading speech-to-text model across Indic languages.
- [Building resiliency in agentic environments](https://hyde.ai/blog/building-resiliency-in-agentic-environments): Research on making agentic systems robust in production.

## Contact and company

- [Get in touch / book a demo](https://calendly.com/asanga-buildwithhyde/new-meeting): Hyde currently collaborates with select enterprise customers; reach out to discuss a partnership.
- [Open roles](https://jobs.ashbyhq.com/hyde): Current job postings.
- [LinkedIn](https://www.linkedin.com/company/buildwithhyde/about/)
- [X / Twitter](https://x.com/BuildwithHyde)

## Optional

- [Privacy Policy](https://hyde.ai/hyde-privacy-policy)
- [Terms of Service](https://hyde.ai/hyde-terms-of-service)
- [Service Level Agreement](https://hyde.ai/hyde-service-level-agreement)