Mariam Mohamed@mariammohamed-m
1
Events attended
1
Submissions made
Egypt
2+ years of experience
About me
I'm Mariam, a Computer Engineering student who loves building things where hardware meets intelligence. I'm most excited by machine learning and reinforcement learning, especially getting models to run on small, constrained devices instead of relying on the cloud. I also enjoy embedded systems, IoT, robotics, and autonomous vehicles, and I've dabbled in quantum computing along the way. I'm a fast learner who likes jumping into unfamiliar problems, and hackathons are my favorite excuse to turn a rough idea into something that actually works.
🤓 Latest Submissions
Relay: Every Stage, All Connected
The problem: Development moves through distinct stages: understanding a codebase, reading a task, planning, implementing, debugging, testing, reviewing, and shipping a PR. Two things make this slow: - Each stage is manually effortful on its own. Reading a messy ticket, tracing a bug, writing tests, drafting a PR are real work every time. - Every stage starts from zero relative to the others. The AI doesn't know the plan. The reviewer doesn't know why a decision was made. A bug fixed once gets fixed again because nothing captured it. Most tools solve one of these problems. I solved both. The solution: Relay is an npm package that turns every stage into a purpose built IBM Bob mode, then connects all of them so context flows automatically between stages and across tasks. -- Relay Onboard explores a codebase in parallel across subsystems and writes a living guide, replacing hours of manual reading. Skip it entirely if you already know the code. -- Relay Brief reads a ticket as PDF or plain text and extracts testable acceptance criteria automatically, flagging anything too vague to verify with a suggested rewrite. -- Plan mode, sharpened by Relay's rules, records not just the chosen approach but every alternative it rejected and why. That context normally dies before review. -- Agent mode implements against the approved plan. -- Relay Debug finds root causes and permanently saves them as reusable lessons the moment they're found, with no manual step required. -- Relay Test generates coverage from the actual requirements, not just the code, and stays honest about what's still untested instead of faking a green checkmark. -- Relay Review runs several checks in parallel: against the plan, the requirements, past lessons, and conventions. -- PR generation assembles a description that already contains the goal, the decisions, the rejected alternatives, the debugging history, and a live merge conflict check. Beyond the stages, Relay also handles the practical realities of real projects: -Multiple tasks at once, each with its own history, tags, and timeline, all visible from one dashboard. - Revisit tracking. If testing sends a task back to implementation or planning, Relay counts the revisit and amends the affected document rather than overwriting it, so every prior version stays inspectable. - Baseline comparison. Relay estimates how long a ticket would take a developer working manually, based on the ticket's actual complexity and historical task data, then compares that against Bob's real time, including any revisits. That gives a concrete, non fabricated number for time saved. - Flexible storage. Relay's context files live in the repo itself by default, versioned alongside the code, so the reasoning behind a decision survives in git history. For teams whose repos grow too large for that, the same context can instead live in a separate linked repo or an external store, with Bob resolving it through the same reference pattern. Target users: Developers and teams on IBM Bob who want every stage of a task easier individually, connected as one workflow, and measurably faster. How they interact with it: Relay installs with one command, npm install -g bob-relay, directly into any existing project. No separate app, no new infrastructure. Developers work in Bob IDE as usual, selecting Relay's modes stage by stage, while a companion dashboard shows the live journey, requirement coverage, the growing knowledge base, and measured time saved. What makes it unique: Most AI tools automate one stage in isolation. Relay automates every stage individually and connects them all, so the ease compounds. The codebase's self knowledge and the developer's speed both improve with every task completed.
27 Sep 2026
