# DataBridge in two pages, for your reviewers Brief for your reviewers · DataBridge by Walnut What it is, the problem it answers, how it works, where it runs, how it was tested, how a pilot runs and who we are. Print it or forward the link. For your IT and security reviewers, the [Technical overview](https://walnut.sg/product/technical/) has the detail. ## [What it is](https://walnut.sg/brief/#what-head) DataBridge by Walnut connects the information spread across your existing systems into one record your teams can use. Ask a question in plain English and get the answer from your data, with the source on every figure. It runs on your own machines, and nothing leaves your control. ## [The problem it answers](https://walnut.sg/brief/#problem-head) - 25% of a team's time goes to searching for answers. Atlassian, State of Teams 2025, 12,000 knowledge workers. - 5.3 hours a week, per person, spent waiting on colleagues for information or recreating work that already exists. Panopto with YouGov, 2018, 1,001 employees. - 42% of an organisation's know-how is held only by individual employees. When they move on, it goes with them. Panopto with YouGov, 2018, 1,001 employees. ## [Two layers on one foundation](https://walnut.sg/brief/#layers-head) A Company Brain that connects your data, and the intelligence layer your teams use to ask questions, get reports and draft routine work. ## [How it works](https://walnut.sg/brief/#how-head) Five steps take your data from the systems it sits in to a record your teams can check. ### Connect Reads databases, documents, spreadsheets, email and the systems your teams use. ### Understand Recognises the same customer, supplier, project or person across every source and links them into one record. ### Convert Gives each job the information in the form it needs: answers, reports, tables for analysis, and data ready for AI tools. ### Prove Each figure read from a file points to the cell, line or page it came from, and shows whether it has been checked. ### Protect Runs on your machines, with the AI model you choose. Each person sees only what their role allows. ## [Where it runs, and who sees what](https://walnut.sg/brief/#security-head) - Your choice of model: use the AI model you prefer, run it on your own machines, or use none. Matching records, reconciling figures, reports and exports work without any AI. - Runs on your machines: installed on hardware you control. When it is installed on your own premises, nothing is uploaded and nothing connects out. - Access by role: each person sees only what their role allows. Confidential material is kept separately, locked with its own key. - A person confirms what AI proposes: facts a model suggests wait for a person to confirm them, and drafted emails go out only after approval. - Every version is kept: corrections add a new version and keep the old one, so you can see what the record said on any past date. - Leave with everything: export the whole record in standard formats, and we restore it into a fresh system in front of you. Your IT, privacy and vendor-risk reviewers receive the full architecture sheet before anything is installed. ## [Test results](https://walnut.sg/brief/#results-head) Each result was measured on synthetic company data with known answers, so it can be checked. A pilot runs the same checks on your files. - 579,140 of 579,140 source links verified. Every link from a figure back to its source file resolved, and none was broken. Synthetic finance data, October 2026. - 368 of 368 figures matched. Answers that combined several systems matched the source database exactly. Synthetic finance data, October 2026. - 0.9 seconds to add a ninth source. No rebuild was needed; 12 of 12 true matches were linked, and 6 of 6 look-alike names were kept apart. Synthetic test data, August 2026. - 26 of 26 planted attacks contained. Instructions hidden in emails and files were quoted as content; in the 26 planted tests, none was followed. Synthetic finance data, August 2026. ## [How a pilot runs](https://walnut.sg/brief/#pilot-head) A pilot takes one workflow your team repeats. We show DataBridge on a sample of your files, agree the scope and measures in writing, install it on your machines and measure against your baseline; then you decide what comes next. It starts with a demo on a sample of your own documents and data. ## [Who we are](https://walnut.sg/brief/#team-head) ### Dev Sirbhaiya Co-founder AI and Automation at Singlife · AI/ML Engineer at Neurowyzr · AI and Automation at Bajaj Finserv · Researcher, NTU-Schaeffler Lab · Tech Lead, NTU AI&ML Club · Director of Projects, MLDA@EEE. ### Mehul Modi Co-founder Production AI agent over 16,000+ records at Changi's AI/ML Lab · Privacy-aware clinical AI at Neurowyzr · Multi-agent AI research at A*STAR. ### Aaruni Garg Co-founder Research Intern at SEC · Architecture Intern at HighSpan, MAK Designs, TAS · Design Intern at WGS. ### Ishaan Sirbhaiya Co-founder Real-time power-grid and industrial cybersecurity research at SUTD's iTrust Centre · Production full-stack systems at Bajaj Finance and across multiple client projects. Walnut is backed by SUTD VIE's Baby Shark Fund. Walnut is based in Singapore. ## [How to reach us](https://walnut.sg/brief/#reach-head) Write to outreach@walnut.sg with one task your team does every week.