Evans Eburu appears to offer specialized engineering services, particularly in cross-border neobanks and stablecoin infrastructures. The mention of founding engineer roles at companies like Figo and Autospend suggests a hands-on approach to building core backend systems. For businesses looking for expertise in agentic trading or payment rails, their track record seems relevant. Is the pricing aligned with the high level of AI safety research they also conduct? It is not clear if their academic work impacts their commercial rates, but the technical capability is evident.
Questions about Evans Eburu
What is Evans Eburu?
Evans Eburu is a software engineer and AI-safety researcher specializing in backend and fintech systems, including payment rails and cross-border settlement. He also conducts independent research on how large language models deceive and whether their private reasoning can be monitored. He is currently an MSc student in Applied IT at Ural Federal University and recently joined Figo as a founding engineer.
What specific benchmarks or tools are available for evaluating AI safety?
The business offers DeceptionBench, a cross-play benchmark for evaluating LLM deception and detection across six frontier models, and MandateBench, which tests if AI payment agents adhere to spending mandates. Both benchmarks are available as preprints on Zenodo with specific DOIs released in 2026.
What are the key performance metrics demonstrated in the DeceptionBench research?
The website highlights a monitor reading only a model's private reasoning that catches a liar, achieving a deception detection AUROC of 1.00 and monitor accuracy of 96% across six frontier models. This research is presented in the DeceptionBench preprint.
Who is the target audience for these AI safety and fintech solutions?
The platform serves individuals and organizations interested in AI safety, fintech reliability, and cross-border payment systems. It is suitable for those needing benchmarks to evaluate LLM deception or verify that AI payment agents stay within signed spending mandates.
What unique features are included in the BOTVBOT project?
The website features a deterministic engine for a project called BOTVBOT, where coaches write plain-English game plans and an AI model picks moves like Jab, Heavy, Block, and Dodge. The system ensures zero randomness so every fight is replayable and verifiable.
How can users access the live demonstrations of the benchmarks?
The website mentions a live demo for DeceptionBench and a preprint on Zenodo for MandateBench. These resources allow users to interact with the benchmarks and read the detailed findings online.
How long has the business been operating and what are the key milestones?
The business has been active since at least 2022, with a personal post from that year. The founder began their MSc in Applied IT in 2025 and released DeceptionBench and MandateBench in 2026.
What specific areas of fintech does the business specialize in?
The business focuses on reliability-critical systems, specifically payment rails and cross-border settlement. The founder has six years of experience building these systems before transitioning to AI safety research.
How can visitors contact the business or find more information?
The website provides links to the founder's email, GitHub, LinkedIn, Scholar, and ORCID profiles. These are the primary methods for contacting the business or accessing their research publications.
What recent publications have been released by the business?
The business released a preprint on Zenodo in 2026 regarding MandateBench, focusing on mandate faithfulness and pre-signature monitorability for agentic-payment LLMs. This work complements the earlier DeceptionBench release.
What is the official website of Evans Eburu?
The official website of Evans Eburu is evanseburu.com.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
The site shows serious tech chops in AI safety and fintech. It’s clear they know their stuff, but it’s not totally obvious how that translates to a standard client project. Seems like a mix of research and engineering.
Accessing their technical resources and live demos is straightforward, making their expertise on AI safety easily available for review.
The profile of Evans Eburu presents an interesting blend of practical engineering and academic research. With six years of experience building reliability-critical systems, including payment rails and agentic trading, the team has demonstrated a strong capability in fintech infrastructure. Their recent work on DeceptionBench, which measures deception and detection in large language models, shows a forward-thinking approach to AI safety. The fact that they are also an MSc student in Applied IT at Ural Federal University suggests a commitment to staying at the forefront of technological developments. The live demo and preprint availability indicate transparency in their methodologies. However, the transition from founding engineer roles to independent research might mean that commercial clients need to ensure their specific needs are prioritized over experimental projects. The technical depth is undeniable, but the service offering seems heavily influenced by the founder’s dual interests in engineering and safety research.
Evans Eburu offers a solid foundation for reliability-critical systems, particularly in fintech and cross-border settlements. The team’s background in building payment rails is evident in their work. However, one might question whether the sheer volume of independent research on AI deception, such as DeceptionBench, dilutes their focus on core engineering deliverables. It seems they are stretching their expertise across both traditional backend infrastructure and cutting-edge AI safety. For someone needing robust, proven systems, their experience is strong, but the breadth of their current projects might suggest a lack of deep specialization in just one area. Still, the results speak for themselves.
Questions & Answers
No questions yet — be the first to ask.
Sign in to post
Your text is saved — you'll come right back to post.
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
About Evans Eburu
Evans Eburu is a software engineer with a focus on backend and fintech systems, including payment rails, cross-border settlement, and agentic trading. He also conducts independent research on AI safety, specifically studying how large language models deceive, how deception can be detected, and whether a model's private reasoning can be monitored. He is the author of DeceptionBench, a cross-play benchmark for evaluating LLM deception, detection, and monitorability across frontier models, as well as MandateBench, a benchmark for evaluating whether AI payment agents adhere to spending mandates. He is currently pursuing an MSc in Applied IT at Ural Federal University and has recently joined Figo as a founding engineer working on cross-border neobank backend infrastructure.
- Website
- evanseburu.com
- [email protected]

