Products in AI & Data

The legal landscape in Ethiopia is constantly evolving, making it difficult for anyone to stay updated on the latest proclamations and cassation rulings. Negarit solves this with a 100% AI-based research tool powered by a vast, ever-expanding database. By automating the search across a massive corpus of laws, it provides students, citizens, and lawyers with instant, accurate answers from a source that never goes out of date.

Currently, 90% of Ethiopian restaurants rely on outdated legacy systems or manual pen-and-paper because modern hardware is too expensive. This causes slow table turnover, lost revenue, and major language barriers between front-of-house staff and the kitchen. TableTap solves this by operating entirely in the cloud. It is a 100% hardware-free order management platform where waitstaff use their own smartphones to process orders in native Amharic. These commands are instantly sent to a digital Kitchen Display System, effectively acting as an autonomous digital employee that speeds up service and eliminates order errors without the need for a single piece of proprietary hardware.

A few AI efforts exist for Ethiopian languages (Amharic, Afaan Oromo, Tigrinya, Somali, Sidamo, Wolaytta, and more), but coverage is thin, and most of the 80+ languages are still left out, there's not enough open, high-quality data to build competitive AI for these communities. How: Dataset.ET builds open-source speech, text, and translation datasets across these languages through a community-driven pipeline a Telegram bot collects real contributor audio, with vote-consensus and honeypot QC to keep quality high. Datasets are hosted on Hugging Face, fully open, so more builders can create Amharic-speaking AI and beyond, closing the long-tail gap others haven't reached.

Dala Studio is the only platform where you can build apps, games, comics, music, videos, marketing campaigns, and AI agents. All from one account. No coding. No jumping between tools. We're not promising "easy." We're promising possible — for Africans, by Africans. So if you're tired of global tools that don't understand your language, your payments, or your market? Click. If you want to stop consuming AI and start building with it? Click. If you just want to see what happens when 105,000 African creators get access to the same AI power as Silicon Valley? Click. One line promise: One account. Unlimited creation.

EHUD AI is an all in one creative platform that helps filmmakers, artists, architects, and businesses turn ideas into cinematic AI generated content. From scriptwriting and image generation to video production and storytelling, EHUD AI makes advanced creative tools accessible to everyone. Our mission is to empower African and global creators to produce Hollywood level visuals, films, marketing content, and digital experiences faster, more affordably, and without technical barriers.

While you were offline, AI reshaped the world. We caught every move — so you don't have to. 2 minutes. Every day. On your schedule.

Every AI tells you you're right. That's the bug. If you build alone, you know the failure mode: you ask ChatGPT whether your plan is good, and it says yes. It always says yes. Nobody pushes back before you spend a month going the wrong way. LiveCrew is an executive team that argues back. Four persistent AI executives — CTO, CMO, Head of Product, Head of Ops — that you run like an actual team: all-hands meetings, 1:1s, and real deliverables at the end instead of a chat transcript. The part that makes it work is memory with receipts. They keep every decision, commitment and number you give them, with the verbatim quote attached. Contradict yourself in March, get quoted from January. Tell your CTO you'll ship by Friday and he asks about it on Friday. Why now: models are finally good enough to do real work. They're just trained to agree with you. The missing piece is structure and memory, not a smarter model. I built this because I needed it. I'm a solo founder with nobody to argue with, and the team runs my company for real — my AI CMO writes LiveCrew's LinkedIn, and before launch my own executives raised five objections to launching. I fixed three of them before going live. Beta is $29/mo, locked at that price, 7-day trial.

AI agents lack the proper grounding and often hallucinate, they also lack the proper tools to conduct deep research. We're very excited to launch the long awaited ScholarXIV MCP today. This MCP provides >30 tools and connects your AI agents and clients directly to ScholarXIV giving it access to our full suite of research tools. Features • Search papers by title, author, abstract, category, or keyword • Read full metadata and abstracts • Bookmark, like, and comment on papers • Create and manage paper collections • Browse and join public collections from other researchers • Create and manage your ScholarXIV Ch.

it helps you figure out which AI tool to use for a task, instead of guessing or Googling around a crowded market.

Social feeds are packed with AI-generated posts. People still want useful updates from humans they follow, but they spend attention sorting synthetic filler. Cloud AI-blockers ask you to trust another server with the feed itself.

The problem: Giving AI agents real tool access (delete records, move money, send messages, deploy code) means trusting them with a blank check. Traditional authorization — OPA, Casbin, AWS Verified Permissions, Okta — only answers "is this agent allowed to do X?" They can't catch the dangerous case: a legitimately authorized agent that's been manipulated into doing the wrong thing — through prompt injection in the data it's processing, a low-confidence/hallucinated decision, or anomalous behavior. "Authorized but wrong" is exactly the failure mode that makes agent autonomy risky in production. How Lelu solves it: Every agent action flows through one authorization call that returns one of four outcomes instead of just allow/deny: allow — proceed deny — blocked human_review — pause, a human approves, the agent resumes compute — reroute to a safer/sandboxed alternative Behind that call is a layered pipeline: a prompt-injection filter → a confidence gate (using verified LLM token log-probs where available, failing closed when they aren't) → policy evaluation (YAML or OPA/Rego) → a risk score (criticality × (1 − confidence) × reliability × anomaly_factor) → a most-restrictive-wins merge so any single layer can only make a decision stricter, never looser → an audit log of every decision. Around it sits agent identity (RS256 workload JWTs / OIDC), an MCP OAuth 2.1 server, and an encrypted OAuth token vault. The result: every action is checked, every decision is logged, and a human is pulled in exactly when it matters — without changing how you build the agent (one SDK call).

Ethiopia’s financial companies, capital markets, and listed stock prices are severely fragmented. We solve this core problem with an AI you can chat with to instantly access unified market data. Our site also features: Data-backed news Marketplace to find investors, partners, or sell businesses Directory of 30,000+ Ethiopian B2B leads with sector filters. We eliminate data fragmentation by centralizing Ethiopian capital market data and company profiles into one unified platform. How we do it: AI Chat: Instantly query scattered financial data and stock prices. Deal Marketplace: Find investors or buy/sell businesses in a single hub. B2B Directory: Access 30,000+ business leads with sector filters, ending the need to go multiple sources. Everything is aggregated, structured, and easy to access.

BugSight is an AI-powered QA assistant that transforms screenshots of UI bugs into developer-ready bug reports in seconds. It automatically detects issues, extracts error details, assigns severity and priority, suggests possible root causes, and generates structured reports that can be exported to tools like Jira or GitHub. This saves QA teams time, improves report quality, and speeds up bug resolution.

Building a startup in Ethiopia is uniquely challenging due to a critical Information Gap." Founders often struggle with three major obstacles: 1. Regulatory Complexity: Navigating National Bank of Ethiopia (NBE) directives, sector-specific proclamations, and the new Startup Proclamation is often a manual, confusing process that requires expensive legal counsel. 2. Lack of Localized Data: Generic AI models and global benchmarks do not reflect the Ethiopian reality—from Akaki warehouse rents and Bole office costs to local smartphone penetration and mobile money adoption rates. 3. Execution Uncertainty: Founders often lack a "high-conviction" roadmap, leading to wasted capital on business models that aren't legally viable or operationally grounded in the local context. JemirUp is a specialized intelligence platform designed to bridge this gap. It solves these problems through three core pillars: 1. High-Conviction Startup Validation: JemirUp doesn't just give generic advice. It uses a proprietary scoring engine to analyze a startup idea across five dimensions: Market Viability, Regulatory Compliance, Unit Economics, Innovation Angle, and Execution Complexity. It provides founders with a deep-dive analysis, identifying "red flags" early—such as foreign investment restrictions or minimum capital requirements—before they spend a single Birr. 2. The Research Studio (Localized RAG): Powered by advanced "Retrieval-Augmented Generation" (RAG), our Research Studio is seeded with a verified library of Ethiopian policies, strategies, rule and regulations, directives, and development plans (like Digital Ethiopia 2030). 3. Localized Benchmarking & Unit Economics: The platform is grounded in real-world Ethiopian numbers. It provides pre-seeded benchmarks for How it Works: JemirUp utilizes a sophisticated "Multi-Agent" architecture. When a user submits a query, an orchestrator routes the task to specialized AI agents. One agent might scan the regulatory library for legal risks, while another pulls market benchmarks for financial modeling. A final "Verification Engine" then checks the response against our "Ground Truth" data to ensure the advice is accurate and grounded in the Ethiopian context.