Doubleword
Enterprise LLM inference stack from the team behind Titan Takeoff.
About
TitanML was founded in London in 2021 out of UCL deep-learning research by Dr. James Dborin, Dr. Fergus Finn, and Meryem Arik. It raised a $2.8M pre-seed round led by Octopus Ventures in October 2023 alongside the launch of Titan Takeoff, an inference server that automated model compression and optimization so teams could deploy LLMs without hand-tuning quantization. The company reported partnerships with Intel and AWS and claimed up to 90% compute-cost reductions in enterprise pilots.
The company has since rebranded to Doubleword, with titanml.co now presenting Doubleword model serving. The current lineup is the Takeoff Inference Server plus a Takeoff RAG Engine for privately hosted enterprise RAG, and the old open-source Takeoff Community Edition on GitHub sits in a legacy phase with no further maintenance. The vendor also publishes a library of 50+ quantized open-source models on Hugging Face under the TitanML org. Evaluate it as a commercial inference stack for regulated or cost-sensitive deployments, not as an open-source project.
What it can do
- Takeoff Inference Server for optimized LLM serving
- INT4/AWQ quantization and multi-GPU orchestration
- Takeoff RAG Engine for privately hosted enterprise RAG
- OpenAI-compatible API endpoints
- 50+ pre-quantized open-source models on Hugging Face
Setup
- Contact the Doubleword team via titanml.co for Takeoff Inference Server access (enterprise).
- Pull the Takeoff container and point it at your model weights (Hugging Face or local).
- Configure quantization and multi-GPU settings for your hardware.
- Serve through the OpenAI-compatible endpoint and monitor with the built-in metrics.
Prerequisites
Enterprise contract via titanml.co; GPU or CPU infrastructure for self-hosted inference.
Use cases
Works with
Trust score — 6/10
Scored 6/10 (verified tier, metrics-v1). Evidence: $2.8M pre-seed led by Octopus Ventures reported Oct 2023 (Entrepreneur Tribune); founded 2021 London from UCL research; titanml.co verified live, now branded Doubleword; HF org TitanML hosts 50+ quantized models. Strengths: funded team, Intel/AWS partnerships reported, real enterprise inference product. Docked for: (1) open-source Community Edition moved to legacy phase with the repo archived; (2) cost-reduction claims (90%, 20x latency) are vendor-reported, not independently verified. License: proprietary.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 5.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 4.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · $2.8M pre-seed led by Octopus Ventures, Oct 2023 (Entrepreneur Tribune)
- · Founded 2021 London by Dborin, Finn, Arik from UCL research
- · titanml.co verified live, now branded Doubleword (model serving)
- · HF org TitanML hosts 50+ quantized models
- · Open-source Takeoff Community Edition in legacy phase; titanml/takeoff repo archived
- · 90% cost reduction / 20x latency claims are vendor-reported
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.