Portkey.ai raises $3m to help teams build AI apps faster

(From left) Rohit Agarwal and Ayush Garg, co-founders of Portkey.ai / Photo credit: Portkey.ai

Today, a proof of concept (PoC) for an AI product integration can be done in two days. However, getting that product ready for the end user could take as long as a year, said Portkey.ai founder Rohit Agarwal.

“A demo-worthy product versus a production-level product are two very different things,” he told Tech in Asia.

A major obstacle companies face while integrating generative AI into their products is a lack of feedback from the end user. This should come after the PoC stage and cover customer usage and costs, as well as make sure the AI model isn’t hallucinating.

That challenge led Agarwal and co-founder Ayush Garg to found the India-based generative AI firm. Both previously worked together at content marketplace Pepper Content, where they created an AI writing assistant and other AI products built on top of large language models (LLMs).

Funding details

• Funding amount: US$3 million
• Lead investors: Lightspeed
• Other investors: Angel investors and individuals from Amazon Web Services, OpenAI, Cloudfare, Postman, and Asana
• Stage: Seed
• Source

Founded in January 2023, Portkey.ai builds tools that allow businesses to monitor their language model operations (LLMOps), connect with multiple LLMs, as well as experiment, improve, and manage prompts. It also offers integrations with LLMs of companies such as OpenAI, Anthropic, LangChain, and LlamaIndex.

The company has raised US$3 million in a seed round led by Lightspeed. Portkey.ai will use the fresh funding to scale the business while building new product capabilities.

Agarwal said his platform aims to help enterprises streamline and save costs while building genAI features. For example, an enterprise could save over 25% in budget spending with smart caching and monitoring for accuracy and latency.

The smart caching system allows clients to send fewer new requests to an LLM model if end users ask similar questions to a genAI bot. This is done by saving the answers to commonly asked questions in the system.

The company currently charges a fee of about US$50 per one million genAI requests. This is in addition to the subscription fees for closed-source LLMs like OpenAI.

Agarwal said he wants Portkey.ai to be a DevOps platform for companies. Building such a platform in-house would cost a company millions of dollars to get the right AI research teams, engineers, and infrastructure.

The firm also offers a security feature that enables enterprises to redact sensitive information when making requests to LLMs.