Gerald Bustilla@Zrald

SpeakerScope answers the question flat transcripts can't: who said what, while they are saying it. The problem: real-time speech-to-text gives you the words, but without speaker attribution a meeting transcript is unusable for recaps, QA, or action items. Cloud diarization services exist, but they post-process audio after upload, bill per minute, and require raw audio to leave the device — a non-starter for privacy-sensitive meetings. The solution: SpeakerScope is a Windows desktop application written in native C++20 that runs two engines in parallel on one audio stream. AssemblyAI Universal Streaming (v3 WebSocket, 16 kHz PCM, word-level timestamps) provides best-in-class transcription. NVIDIA Nemotron-3-Diarization runs entirely on-device through the audio.cpp GGUF runtime — no GPU, no Python — labeling up to 8 concurrent speakers with live open-turn streaming. A custom fusion engine aligns both streams in the timestamp domain: every AssemblyAI word is attributed to the speaker channel active at its midpoint, producing a live, color-coded transcript. Users see animated speaker lanes (Person 1–8), can rename speakers, filter the transcript per person, and export per-speaker or combined transcripts. Engineering highlights: we patched audio.cpp's streaming decoder to emit open (still-growing) turns so labels appear while people talk, not only after pauses; tuned the chunk geometry to reach 7.6x realtime on a 12-core CPU; and designed a GlobalSpeakerResolver layer (voice-embedding clustering) that extends the 8-channel concurrency limit to unlimited total speakers. Verified: 117/117 automated tests, plus a headless end-to-end simulation streaming a 55.8s 4-voice meeting through real AssemblyAI and Nemotron inference, producing a correctly attributed, overlap-flagged transcript. Stack: C++20, CMake/vcpkg, Dear ImGui, miniaudio, IXWebSocket, AssemblyAI Streaming v3, audio.cpp (GGUF), Nemotron-3-Diarization, GoogleTest.

## The Problem Modern developers don't use one AI assistant — they juggle several at once. IBM Bob lives in its IDE, GitHub Copilot lives in VS Code, and terminal agents (Gemini CLI, Claude Code, Codex) live in shells. Each one is powerful alone, but the work *between* them is entirely manual: - **Onboarding is repeated.** Every agent must be told the project layout, conventions, and current task separately. A new repository can take hours of back-and-forth before any agent is productive. - **Parallel work is copy-pasted.** Splitting a feature into "Bob writes the HTML, Copilot writes the CSS, the terminal agent writes the docs" means a developer manually writes three prompts, watches three windows, and reconciles three answers by hand. - **Status is invisible.** Once a task is handed to an agent, there is no single place to see who is done, who is still thinking, and who silently failed — the developer polls each window. - **Results never come back.** Prompts go out; answers stay trapped in each tool's chat panel. Nothing flows back into one report. - **Release hygiene is forgotten.** API keys and credentials leak into commits because scanning before a push is yet another manual step. The net effect: the developer becomes a human message bus between AI tools — constant context-switching, duplicated instructions, missed failures, and rework.

B Scout — A Multi-Agent Expansion Intelligence Platform B Scout solves this by deploying four specialized AI agents that collaborate through Band's shared interaction layer. A user simply submits a business brief — for example, "Open a specialty coffee shop in Naga City, Philippines" — and the agents take over: 1.Orchestrator (Expansion Coordinator)— Receives the brief through a Band room, builds shared market context, discovers available peer agents using Band's participant tools, and recruits the right specialists by capability. 2. Location Scout (Real Estate Analyst) 3.Competitor Analyst (Market Saturation Specialist) 4.Business Planner (Strategy and Approval Packager)

The application automates deep-dive web research and compliance profiling using the following core capabilities: Active Real-Time Scraping: Queries live web data using standard search pipelines or tunnels requests through robust Bright Data SERP/Web Scraping APIs to gather up-to-the-minute records across six key sector categories: Jobs & Hiring, Corporate News, SEC & Compliance Filings, Workplace Reviews, Official Domain Health, and Social Sentiment. Specialist Multi-Agent Pipeline: Deploying custom-modeled cognitive agents (such as the Lead Orchestrator, Financial Solvency Expert, Regulatory Counsel General, Workplace Operations Officer, Public Relations Sentinel, and Social Pulse Expert) to analyze, sanitize, and verify scraped sources under strict category filters. Cognitive Signal Classification: Dynamically processes individual research entries to extract actionable "Pulse Signals," automatically sizing positive developments and negative anomalies on a normalized severity scale. Risk Metrics & Executive Reporting: Synthesizes the cumulative vector telemetry to compute overall risk scores and compiles detailed multi-sector advisory audits ready for threat mitigation reviews.

Bob is a high-performance, local-first voice assistant designed to bridge the gap between human speech and operating system automation. Built using the modern Tauri framework, Rust, and React, Bob delivers sub-second response times without the overhead of heavy native frameworks. At its core, Bob features a dual-engine architecture: Super-Low Latency Speech Recognition: Powered by Deepgram’s real-time streaming API, Bob actively listens for the wake word ('Bob', 'Bub') and continuously transcribes audio in real-time, displaying live interim feedback. Hybrid Command Parsing: Commands are analyzed using Gemini AI for complex, conversational intents, with a robust Local NLP Parser fallback. The local parser instantly strips conversational filler words ('can you please', 'so', 'but') for reliable offline execution. Bob’s capabilities span across three major categories: System & OS Automation: Launch any application (Notepad, VSCode, Chrome), navigate system folders (Documents, Downloads), manage active windows, or fetch system telemetry (disk space, battery life). Native Browser Control: Browse completely hands-free. Command Bob to open specific URLs, scroll up or down, or simulate mouse clicks directly through low-level Windows APIs. Continuous Dictation & OS Shortcuts: Use phrases like 'transcribe' or simply speak naturally—Bob's intelligent fallback automatically types dictated text character-by-character into whatever field or document has the active cursor focus. You can also trigger system hotkeys, take screenshots (via the Snipping Tool), and execute copy/paste actions purely through voice commands.

Autonomous Intelligent-Retail-Hub is a state-of-the-art e-commerce platform and autonomous back-office operations console. Designed to demonstrate the future of automated commerce, the project features a dual-interface system. On the storefront, buyers interact with a fully voice-enabled shopping assistant powered by low-latency Speechmatics real-time transcription and fluid synthesis. Behind the scenes, a sophisticated multi-agent orchestration team (Main, Sales, Checkout, and Eye agents) coordinates search, cart management, sequential checkout flows, and input-sanitization guards. Meanwhile, the back-office runs a continuous 8-second self-healing autonomy loop. When simulated operational disruptions occur—such as shipping carrier delays or inventory stockouts—autonomous agents immediately intervene. The StocksAgent replenishes inventory, the ShippingAgent adjusts carrier routes to bypass delays, the RefundsAgent issues goodwill vouchers, and the TicketsAgent utilizes Qdrant Vector DB with Featherless embeddings to perform RAG semantic lookups and automatically answer buyer tickets with verbatim source citations. Operators monitor the entire ecosystem through a unified dashboard with sub-second Server-Sent Events (SSE) telemetry, chaos injectors, and dynamic model-switching controls.

The Catog Intelligent Enterprise Solution solves the "Black Box" problem of AI agents in highly regulated industries. While agentic workflows offer immense productivity gains, enterprise security teams often block deployment due to concerns over prompt injection, PII exfiltration, and lack of auditability. CATOG bridges this gap by building a "Trust Layer" where security is a first-class citizen, not an afterthought. CATOG doesn’t just analyze data; it provides a comprehensive governance dashboard and regulator-ready audit trails. Every decision is backed by an audit hash for chain of custody, ensuring that high-stakes legal automation remains secure, compliant, and fully auditable for industries like finance, healthcare, and legal operations. By using Lobster Trap as the "floor" and Gemini as the "brain," CATOG defines the next generation of intelligent, secure enterprise solutions.

Catog Automation is a state-of-the-art, AI-orchestrated platform designed to bridge the gap between human intent and complex digital workflows. Unlike traditional RPA tools that rely on brittle selectors and hardcoded scripts, Catog leverages advanced Vision-Language Models (VLM) to understand and interact with any application—browser or desktop—exactly as a human would. At its core, the system utilizes a high-performance serving stack optimized for AMD ROCm and AMD Instinct™ MI300X hardware. By orchestrating a specialized multi-model pipeline—utilizing Qwen-VL for UI perception, OmniParse for structured data extraction, and Qwen-Coder for real-time execution—Catog "sees" the environment, identifies interactive elements, and generates precise automation patches on the fly. Key Features & Updates: OmniParse Integration: Advanced data ingestion that converts complex UI screenshots and documents into structured, LLM-ready context. Cross-Platform Support: Full compatibility with macOS and Windows desktop environments. Self-Evolving Intelligence: Integrated self-learning capabilities that allow the agent to adapt to UI changes and refine execution logic autonomously.

Six Matches. One leaderboard. Build on AMD, earn points and compete for prizes.

🌎 Hybrid Hackathon | 💻 Online Build: 3–8 November 2026 | On-site Phase: 8 November 2026 📍 On-site Venue: Pierpont Place 163 W Pierpont Ave, Salt Lake City, UT 84101, USA ✨ Please note that travel and accommodation are not covered.

🌎 Hybrid Hackathon | 💻 Online Build: October 16–19, 2026 | 📍 On-site: October 19–20, 2026 The official hackathon of AI & Big Data Expo Europe, organized by TechEx Events and lablab.ai. Build AI applications and agents across AI, cybersecurity, cloud, IoT, automation, edge computing, digital transformation, and data infrastructure, then showcase your project live at one of Europe's largest enterprise technology events. 📍 On-site Venue (October 19–20): RAI Amsterdam Europaplein 24, 1078 GZ Amsterdam, Netherlands 📍 If approved for the on-site experience, you'll also have the opportunity to showcase your project live. Travel and accommodation will not be covered.

🌎 Hybrid Hackathon | 💻 Online Build: 12-17 October 2026 | On-site phase: 17-18 October 2026 To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. The AMD Developer Hackathon: ACT III is a hands-on event for developers, founders, engineers and builders who want to push what’s possible with AI on real infrastructure. 📍 If approved for the on-site experience, you'll also have the opportunity to showcase your project live. Travel and accommodation will not be covered.

🌎 Online Hackathon | 💻 Online Build: Sept 26 - Oct 5, 2026 A virtual, global hackathon. Build from anywhere in the world. Dark Factory is the official hackathon of WeAreDevelopers World Congress North America, presented by BAND, but you don't need to be in San Jose to take part. For one week, you build a software factory in BAND Desktop: a band of coding agents that plans work, implements it and checks its own results. 🎟️ Bonus for sign-ups: you are eligible for a free ticket to WeAreDevelopers World Congress North America in San Jose (Sep 23–25). If you do go, stop by the BAND booth: a surprise is waiting for you.