Hamza Ali@hamza____ali

Trade Verification Firewall is an AI-native options trading agent built for the Alpaca AI Trading Agents Hackathon that solves a core problem with autonomous trading bots: blind trust in a single AI's judgment. Instead of letting one model generate and execute trades, we split the decision into two independent AI checkpoints. The Alpaca API streams live price and options data into a Gemini-powered Signal Agent, which analyzes the market and proposes a trade. That signal doesn't go straight to execution — it passes to a second, independent Gemini Verifier Agent whose only job is to critically review the first agent's reasoning and either approve or reject it. Only signals that clear both checkpoints are sent back to Alpaca to place a paper trade. Every step of this process — the original signal, the verifier's reasoning, the final decision, and the trade outcome — is logged to Supabase and Google Sheets in real time, creating a complete, timestamped audit trail. This makes the system's decision-making fully transparent and reviewable, rather than a black box. n8n orchestrates the entire pipeline, connecting Alpaca's API, the two Gemini agents, and the logging layer into a single automated workflow that's easy to test, debug, and extend. The result is a trading agent architecture built around a simple principle: no trade fires on a single AI opinion.

KnowFlow AI is an AI workplace assistant that transforms static, dense documents into interactive knowledge sources using Retrieval-Augmented Generation (RAG). People often lose hours manually searching through PDFs, manuals, and reports, and standard AI chatbots risk hallucinating facts when they lack real document grounding. KnowFlow AI solves this by letting users upload PDF or text documents, which are automatically parsed, chunked, embedded, and indexed for fast semantic search. Users can then ask natural-language questions and receive concise, accurate answers with citations pointing directly to the source document and excerpt, so every answer is verifiable. The app features secure authentication with per-user data isolation, a document management dashboard with processing status tracking, persistent chat history, and a clean, responsive UI. Under the hood, the frontend is built with React, TypeScript, and Vite, styled with Tailwind CSS and shadcn/ui. The backend runs entirely on Supabase — using Postgres with pgvector for vector storage and semantic search, Supabase Storage for document files, Supabase Auth for secure login, and Supabase Edge Functions for server-side processing. Embeddings and chat completions are powered by the AIML API (OpenAI-compatible), using text-embedding-3-small and GPT-4o-mini. The result is a fast, secure, and genuinely useful tool for anyone who needs instant, trustworthy answers from their own documents.

72% of diabetics are only diagnosed after a complication has already appeared, and damage to the kidneys, nerves, eyes, or heart can develop silently for 4-7 years before that. Today's screening checks one specialist, one chart, one metric at a time. GlycoSwarm AI fixes that with a 4-agent specialist pipeline: dedicated Renal, Neuropathy, Retinal, and Cardiovascular agents each reason over a different slice of a patient's real lab panel, then a Synthesis Agent weighs all four findings against each other and outputs one ranked referral recommendation, all generated live, not pre-computed or hardcoded. Every specialist writes and executes its own Python risk-scoring code against the patient's actual labs in real time, visible in a live reasoning terminal. The system runs on genuine on-GPU inference: Gemma 4 26B served via Ollama on an AMD Instinct MI300X (AMD Developer Cloud), with Fireworks GLM 5.2 as a real, user-triggered fallback. If neither provider is reachable, it honestly reports "unavailable" rather than faking a clean result. We validated the full pipeline across 104 real, de-identified CDC NHANES patients, with 90 flagged in at least one organ system. The dashboard (Next.js, FastAPI, LangGraph) renders an anatomical risk map, live agent terminal, and an exportable clinical brief. It's a prototype, not a certified clinical device, but the architecture generalizes beyond diabetes to chronic kidney disease, metabolic syndrome, and post-transplant care, as a SaaS subscription or EHR-integrated license, against a $1T+ global health spend.

🌎 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 | 💻 Dates: 28 August–4 September 2026 | 🏆 $6,000 Prize Pool Build AI trading agents on Alpaca — autonomous agents and trading apps using Alpaca's Trading API, MCP server and CLI. 🌍 Fully online hackathon. Join and build from anywhere in the world.

🌎 Online Hackathon | 📅 August 3–10, 2026| 💻 Powered by native.builder The official native.builder hackathon, hosted by NativelyAI and lablab.ai. Join founders, developers, designers, and builders to create real AI-native applications powered by native.builder: from idea to deployed product, fully in the cloud. 💻 This hackathon is fully online. No travel required.