🤝 Top Collaborators
🤓 Latest Submissions
SiteSpeak
Next-gen hands-free voice safety companion: speak naturally, and an AssemblyAI voice agent activates HAZMAT SDS protocols, plots hazards on a Digital Twin, files certified OSHA-301 reports, and triggers emergency alarms in real-time. In heavy construction, chemical plants, and logistics facilities, workers wear heavy gloves and face high-stress emergencies. Paper logs and radios cause 20+ minute reporting delays—and toxic spills or fires don't wait. SiteSpeak 2.0 is an autonomous industrial voice safety agent built on the AssemblyAI Voice Agent API Universal-3 Pro STT, VAD turn-taking, managed LLM, and bidirectional client-side tool execution). A worker taps one button and describes the situation naturally.
30 Sep 2026
AlphaCondor
Problem & Motivation Retail algorithmic options trading faces two major hurdles: capital inefficiency and risk control. Traditional Cash-Secured Puts (CSPs) on blue chips lock up $30K–$60K in cash collateral per contract for small yields. Furthermore, connecting generative AI directly to broker execution creates severe hallucination and leverage risks. The Solution: AlphaCondor AlphaCondor is an autonomous AI options trading system built on Alpaca's developer platform, operating a dual-strategy options income engine: 1. Core Wheel Strategy: Systematic Cash-Secured Put (CSP) to Covered Call (CC) cycles on blue-chips (AAPL, MSFT, NVDA, SOFI, SPY) capturing steady theta decay. 2. 0DTE Iron Condors (SPY): A high-margin-efficiency defined-risk engine trading 4-leg credit spreads with defined $500 margin per spread instead of $30K+ collateral, unlocking 10x–20x greater capital efficiency. Decoupled Architecture: AI Advisor + Risk Firewall AlphaCondor strictly decouples AI cognition from execution: • Perception: Live prices, historical bars, option chains with Black-Scholes Greeks, and company news via Alpaca's API. • AI Cognition: Google Gemini (gemini-3.5-flash-lite) evaluates trend, volatility, and sentiment into structured Pydantic trade signals. • 7 Deterministic Risk Gates: Mathematical rules (70% confidence floor, 100% cash collateral, concentration caps, covered share backing, and liquidity filters) that no LLM can bypass. • Execution: Limit orders with explicit PositionIntent, WebSocket fill monitoring (TradingStream), Alpaca MCP server, SQLite journal, and Streamlit dashboard. Verified Results (2024 – 2026 Backtest) • Starting Capital: $100,000.00 • Ending Portfolio Equity: $117,347.00 (+17.3% ROI) • Total Premiums Collected: $40,703.00 • Strategy Win Rate: 84.5% across 276 executed trades • Dedicated Competition Account: PA3QPTGUKZ8P
4 Sep 2026
Quadravox
Quadravox is an AI video-to-style captioning agent built for the AMD Developer Hackathon (Track 2). It watches a video clip and generates precise, on-tone captions in four styles — Formal, Sarcastic, Humorous (Tech), and Humorous (Non-Tech) — generalizing across nature, urban, animals, people, sports, food, and tech content. Architecture (high level): Preprocessing — Evenly-spaced keyframes are extracted across the full clip (including its final frames) at a resolution high enough to catch fine details like signage or device type. Audio is pulled in parallel and transcribed for spoken-word context (optional, non-fatal — captioning still works if this step fails). Structured Vision Captioning — Keyframes and transcript are sent to a vision-capable LLM in a single structured call, with a fallback model on standby. JSON mode plus a strict, few-shot prompt returns all four styles as clean JSON, avoiding the chain-of-thought leakage and truncation that break naive prompting. A defensive parser and placeholder guard reject any stray or template output. Self-Evaluation & Repair — Each caption is scored on accuracy and style-match, mirroring the judge's rubric; only weak styles get regenerated, preserving the strong ones. Key features: Specific & factual — all four captions derive from the same high-detail keyframes and are prompted to name concrete specifics, so no style contradicts another. Parallel & fast — clips are processed concurrently, so the full set finishes comfortably within the time limit. Resilient fail-safes — checkpointed, pre-seeded output means results are always valid, complete JSON — no style scores zero by omission. Lightweight — a minimal containerized service reads a task file and writes results, no heavyweight dependencies.
13 Jul 2026
