Muhammad Rehan@muhammad_rehan_417

Voice Action Gate is a security layer for AI voice agents that prevents sensitive actions from being executed based on guessed, missing, or ambiguous intent. It verifies that required parameters were explicitly provided, checks risk and confirmation requirements, and blocks unsafe requests before they reach protected tools. The system separates AI understanding from authorization, maintains an audit trail, and demonstrates secure workflows such as deployments, database actions, and simulated fund transfers. Its goal is simple: AI may understand what you mean, but it should only act on what you clearly authorized. :chatgpt-content-reference{index="0"}

ToolFence is a task-scoped security gateway for AI coding agents, built for the IBM Bob 2.0 Hackathon. Its goal is simple: let an AI agent remain useful without giving it unrestricted access to repositories, CI systems, tickets, deployments, or secrets. Instead of granting broad permissions, ToolFence lets IBM Bob reason about the minimum capabilities required for a task. Bob submits a task-specific permission proposal, while a separately trusted developer-defined approval ceiling remains hidden from the model. ToolFence validates the proposal, activates the resulting policy, and then checks every protected tool call before execution. The MVP exposes 11 protected capabilities across tickets, repositories, pull requests, CI, releases, and secrets, plus 4 ToolFence control-plane tools. The system is designed to fail closed: if there is no active policy, a tool is not granted, a resource is outside scope, or arguments are invalid, the action is denied and not forwarded to the backend. Our golden demo uses the task: “Fix BUG-17, run CI, and create a pull request.” ToolFence grants only 6 of 11 available capabilities. A forbidden secret.read request is denied with TOOL_NOT_GRANTED, while the legitimate workflow succeeds through ticket.get, repo.read, repo.write, ci.run, ci.status, and pull_request.create. ToolFence also includes append-only audit logging, runtime policy snapshots, a read-only FastAPI observability layer, and a routed React security dashboard showing policy state, capability boundaries, activity, audit evidence, and benchmark results. In our replay benchmark, ToolFence matched 16/16 cases, blocked 100% of forbidden actions, achieved 0% false denials, 100% boundary-case accuracy, and reduced the active privilege surface by 45.45%. The core principle is: Bob reasons. ToolFence enforces.

TaskForge VLA is a simulation-first embodied AI system built for the AI Infra Summit Hackathon 2026, targeting the Intel Bimanual VLA Manipulation with Multi-Modal Reasoning track. Users can give natural-language commands such as “Pick both blocks” or “Put the red block near the blue one.” TaskForge interprets the instruction with an OpenRouter-powered LLM, converts it into a typed, structured goal, and falls back to a deterministic parser if needed. The system observes the simulated world, generates a symbolic plan, and uses a reachability-aware bimanual coordinator to decide which arm should perform each action. Before execution, a deterministic validator checks objects, actors, dependencies, grasp states, and supported actions. Validated plans are executed in MuJoCo using deterministic primitives such as pick, place, and move-home, with analytical inverse kinematics controlling the arms. After execution, TaskForge re-observes the world and verifies whether the original goal is satisfied using predicates such as `left_of`, `right_of`, `near`, and `grasped_by`. A key feature is closed-loop recovery. If an action fails, TaskForge preserves completed work, observes the updated world, creates a residual goal for only the unfinished task, replans, validates the new plan, and continues execution. The project also includes a Streamlit dashboard for task input, plans, world state, execution status, recovery events, and final verification, plus a live MuJoCo viewer for visual robot simulation. Core principle: AI interprets. Structured systems validate. Deterministic robotics executes. The environment decides success.

Hedgify is an autonomous, multi-agent portfolio insurance system built for the Alpaca AI Trading Agents Hackathon (lablab.ai). It solves a problem every equity investor knows intimately: the panic-sell. When markets crash, retail investors liquidate their holdings to stop the bleeding — crystallizing losses, triggering tax events, and inevitably missing the rebound. Hedgify replaces that reflex with insurance: instead of selling your stocks when danger approaches, the system automatically buys protective put options — capping your downside while you keep every single share and ride the recovery. The system is architected as a Supervisor–Worker multi-agent pipeline. The Monitor Agent polls live portfolio data from Alpaca's paper trading API, tracks the all-time peak value, and computes real-time drawdown the moment any poll lands. When the drawdown crosses its threshold, the Monitor persists a snapshot, fires a structured alert into the database, broadcasts it over WebSocket to any connected dashboards, and hands off to the Hedge Executor Agent. The Executor then runs its gauntlet: an idempotency guard first checks whether the symbol already carries active protection (sixteen duplicate alerts become exactly zero duplicate orders); the live kill-switch is consulted, letting an operator disable autonomous trading mid-crisis without restarting anything; then the agent queries Alpaca's live options chain over its dedicated data API, matches the target strike (5% out-of-the-money by default) to the closest actually listed contract, validates expiry against the settings table, checks the premium against the configured budget, and places a real order through Alpaca's trading API — then polls asynchronously for the actual fill price, recording genuine market premiums rather than estimates.

LeakLogic AI is an explainable AI business-auditing platform built for small and medium-sized businesses that need to understand where profit is being lost. Users upload sales, refunds, supplier-cost, and inventory CSV files through a modern web dashboard. The FastAPI backend validates and normalizes the data, then runs four deterministic Python and pandas detectors for refund anomalies, discount leakage, supplier margin compression, and inventory drag. Each finding includes the affected entity, measured change, supporting evidence, confidence score, estimated dollar impact, likely cause, and recommended action. Findings are ranked by financial impact so decision-makers can focus on the most important issues first. After the calculations are completed, Fireworks AI transforms the verified structured findings into a concise executive narrative. The model is instructed not to invent figures, causes, evidence, or recommendations. If the AI service is unavailable, LeakLogic AI automatically returns a deterministic fallback summary, ensuring the audit still completes. The application is fully containerized with Docker and Docker Compose. The production frontend is deployed on Netlify, while the FastAPI backend is deployed on Vercel. The project uses Fireworks AI API access from the AMD hackathon ecosystem and was developed for the Unicorn Track as a product-focused AI solution. LeakLogic AI turns disconnected operational data into a clear answer to one business question: where are we quietly losing money, why does it matter, and what should we review next?

💻 Online Hackathon | ⏳ Build: 48 hours | 📅 September 25–27, 2026 The official IBM Bob 2.0 hackathon. Build what's next in AI-assisted development, alongside the developers and experts shaping it. ⭐ Get help from experts throughout the event 🤝 Join a team or build solo 💰 Prize pool: $12,000 🧑🏻‍💻 Register before the Kick-Off Stream to start building with Bob

🌎 Hybrid Hackathon | 💻 Online Build: September 10–16, 2026 | 📍 On-site: September 15–17, 2026 The official hackathon of AI Infra Summit — co-organized by Kisaco Research and lablab.ai. Build production-grade AI applications on the modern AI infrastructure stack, alongside 8,000+ attendees, researchers, founders, and infra leaders. 📍 On-site Venue (September 15–17): Santa Clara Convention Center 5001 Great America Pkwy, Santa Clara, CA 📍 On-site participation is by invitation only. Travel and accommodation expenses will not be covered.

🌎 Online Hackathon | 💻 Month-long challenge | Dates: Sep 1–30, 2026 | 🏆 $10,000 Prize Pool ($5k cash + $5k in AAI credits) A month-long online challenge run by lablab.ai together with AssemblyAI, the Voice AI infrastructure company for builders. Every participant builds on AssemblyAI. You can join at any point — registration stays open for the whole build window, so you can start on day one or pick the project up halfway through and still submit. 🌍 Fully online hackathon. Join and build from anywhere in the world.