The archive · AI & Models · Strategic decision · 2020–2025
NeuralSpace's local-languages-first AI bet shuts down after five years
London language-AI startup bet cloud giants were wrong about low-resource languages; a founder post-mortem details the 2025 wind-down after a $2.8M seed.
NeuralSpace
What the business is
Language AI SaaS: a platform with APIs for machine translation, speech-to-text, text-to-speech, entity recognition and language understanding in more than 100 languages, focused on locally spoken languages.
Starting capital：$2.8M seed investment in 2022, with Techstars among the investors
How it started
Founded by computer-science students Ayushman Dash and Felix Laumann, who met while studying in Europe. They found that NLP models from Google, Microsoft, IBM and AWS were inaccurate for locally spoken languages in real use, so they built NeuralSpace with a local-languages-first approach; the company was headquartered in London with offices in India and Tunisia.
What happened
By September 2022 NeuralSpace had 22 employees, a live platform used by more than 300 companies, clients including conversational-AI firm Hello Ebbot and AI marketplace Eden AI, and a $2.8M seed investment with Techstars. It offered more than 100 languages against roughly 30 for Google, Microsoft Azure and IBM Watson, and competed with Slang Labs, Deepdub, Papercup and Lovo on accuracy and breadth.
How it ended up
In March 2025, co-founder Ayushman Dash announced he was shutting the company down and published a detailed post-mortem about the decision. The team had been reduced to six people, all with the company since the beginning, who stayed to close it down.
Background
NeuralSpace was a London-based language-AI startup founded by computer-science students Ayushman Dash and Felix Laumann. Its bet: Google, Microsoft, IBM and AWS models were inaccurate for locally spoken languages in real-world use, so a startup that trained models with a local-languages-first approach could win developers across India, the Middle East and Southeast Asia.
The company offered a SaaS platform with APIs for machine translation, speech-to-text, text-to-speech, entity recognition, language understanding and transliteration across more than 100 languages. It differentiated on breadth — more than 100 languages versus roughly 30 for the big cloud vendors — and on accuracy for low-resource languages such as Gujarati, Telugu, Gulf Arabic and Oriya.
By September 2022 NeuralSpace had 22 employees, offices in India and Tunisia, more than 300 companies using the product, clients including conversational-AI firm Hello Ebbot and AI marketplace Eden AI, and a $2.8M seed investment with Techstars. It charged pay-as-you-go and competed with Slang Labs, Deepdub, Papercup and Lovo.
In March 2025, co-founder Ayushman Dash announced the company was shutting down and wrote a post-mortem describing the emotional weight of the decision. The team had been reduced to six people, all with the company since the beginning, who stayed until NeuralSpace closed. His essay urged founders to stay grounded, stay honest, and trust reality more than the idea.
What has to be true
- Language AI for low-resource languages is a fragmented market where cloud giants give away basic models as loss leaders.
- Pay-as-you-go API revenue from a few hundred companies could not fund in-house model training across 100+ languages.
- Enterprise buyers in the target regions defaulted to big vendors' bundled platforms, making distribution the binding constraint.
- The founder concluded the honest call was to stop while the core team could still be treated fairly.
What can be applied
Building better models than cloud giants is not enough; the hard problem is distribution, and the hardest founder skill is knowing when to stop and trust reality over the idea.
Aftermath
As of September 2026, NeuralSpace is closed: the founder's March 2025 post-mortem describes an orderly wind-down with a six-person core team, and no later activity or relaunch has been documented in the sources reviewed.
Sources
- How I Felt While Shutting Down My Company, NeuralSpace
- How NeuralSpace helps developers build language AI models in Oriya, German, and many more local languages
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card