Reduce backend latency from a peak of 14,693 ms to a low of 17 ms
Serve 66.67% more requests per second now compared to November 2024
Optimize VM CPU utilization for increased queries and user capacity
Reduce VM CPU Load significantly with further optimization
With the help of Google Cloud, AI Seer is well-equipped for quick development and deployment, delivering solutions with intuitive interfaces that enhance productivity and collaboration for users and internal teams.
With the help of Google Cloud, AI Seer is well-equipped for quick development and deployment, delivering solutions with intuitive interfaces that enhance productivity and collaboration for users and internal teams.
Information has never been more critical to the world today, with data being utilized in different ways to make decisions that have an impact, both big and small. However, there is always the risk of misinformation and inaccurate data being shared among an audience that might not always be in the best position to ascertain its legitimacy. This is where AI Seer comes into the picture.
Founded in 2019, this deep-tech startup is on a mission to check untruths by becoming the filter for malinformation, disinformation, and misinformation, preventing the wrong facts from negatively impacting the world. This takes the form of two key projects.
Multilingual social media AI fact checker Facticity.AI, one of the 14 winners of TIME’s Best Inventions of 2024 in the AI Category, is designed to prioritize explainability from credible sources. Compared to other Large Language Models (LLMs) search applications, the platform helps to fact-check anything seen or heard, powered by a new level of transparency about statement-level references, enabling it to outperform other LLM-search apps.
AI Seer has also invented the world’s first Multi-Spectral Reality Detector (MSRD) to enhance the discernment of intention, emotion, and (re)cognition, conducting over 300 experimental sessions as part of its refining process. The MSRD has been a breakthrough solution in cognitive and autonomic nervous system monitoring and credibility assessments, drawing interest from those seeking to decentralize truth detection and fact-checking.
As the work continued, it became clear to the AI Seer team that further expansion and refinements of their solutions required further optimization of the increasing workloads and the applications needed to deliver better performance. With Google Cloud being a platform with different solutions that could help and was easy to use, it became the top choice for AI Seer moving forward.
Impactful changes that make a difference
With a small but dedicated team, it became necessary to leverage technological solutions as much as possible to improve working processes. Using Cloud Run and Artifact Registry to house docker containers allows the fact-checking capabilities to perform at their best – even when there is a surge in traffic and queries – with auto-scaling and possible quick switches doing the heavy lifting. Compared to November 2024, Facticity.AI can serve 66.67% more requests per second.
On the same front, the increasing number of API calls and queries benefits from the implementation of both Gemini and Vertex AI within the overall system, improving the user experience and bringing down the overall costs. Latency has also dropped from a peak of 14,693 ms to an impressive low of 17 ms.
Overall, the efficiency yielded by the Google Cloud platform has allowed AI Seer to significantly shorten development cycles, enabling faster rollouts of new features and updates. In terms of the time-to-market for solutions, the results have never been better.
“Our biggest concern with adopting generative AI is addressing hallucinations. To mitigate this, we’ve integrated a multi-LLM approach to cross-validate outputs and reduce reliance on a single model. Continuous monitoring and feedback loops further help refine outputs, significantly minimizing hallucinations and improving reliability in delivering accurate AI solutions,” says Dennis Yap, Founder of AI Seer.
Our biggest concern with adopting generative AI is addressing hallucinations. To mitigate this, we’ve integrated a multi-LLM approach to cross-validate outputs and reduce reliance on a single model. Continuous monitoring and feedback loops further help refine outputs, significantly minimizing hallucinations and improving reliability in delivering accurate AI solutions.
Dennis Yap
Founder, AI Seer
The platform’s ease of use has also been critical to the team, helping to reduce the learning curve, boost overall productivity, and drive increased collaboration. Whether users are technical or non-technical, there is always a way to contribute, streamlining workflows that draw the focus back to delivering the best solutions for the audience.
“Our collaboration with Google has been instrumental in enabling us to expand Facticity's capabilities. This includes releasing a Word Add-In and a Chrome extension, which is in the works, which will further enhance user accessibility and productivity,” says Yap.
“Google's ecosystem has not only facilitated rapid development but also empowered us to stay ahead in the generative AI space, aligning with our vision of creating the world’s leading automated fact-checking tool.”
The positive impact brought by working on the platform has further emboldened the team to aim higher. While AI Seer plans to continue using Google Cloud solutions for hosting virtual machines and docker containers, benefiting from the reliability, performance, and scalability of the infrastructure, the team is also looking forward to exploring updated products such as Gemini 2.0 Flash.
AI Seer is a deep-tech startup on a mission to enable more perceptive people and machines by building authentication into information streams to deter untruths.
Industry: Technology
Location: Singapore
Products: Cloud Run, Artifact Registry, Gemini, Vertex AI