
1
1
3+ years of experience
Hey, I'm Thapelo; a Financial Wealth Planner and systems developer from Johannesburg, South Africa. By day I helping individuals, families, and businesses build real, lasting wealth. But I also build things, like, actually build them. I coded my own personal website from scratch, and I've been getting deep into AI projects that sit right at the intersection of finance and technology. That's the space I find most exciting. I genuinely believe AI can change how people across the world access financial guidance, not just the privileged few, but everyday people who deserve the same quality of advice. And I don't just want to talk about it, I want to build it. I'm here to collaborate, ship ideas fast, and work with people who are just as obsessed with using AI to solve real problems. If you're into financial inclusion, economic empowerment, or just making complex things more accessible, we'll get along just fine.

π€ NEXUS Trading AI - Model Description System Overview NEXUS is a self-improving multi-agent trading system operating without GPUs or gradient descent. It uses a reputation-weighted voting mechanism where four agents compete, adjusting their influence based on real PnL to find dominant strategies within 24β72 hours. The Four Agents 1. MomentumAgent: Uses RSI to gauge trend strength. Thrives in trending markets. 2. SentimentAgent: Blends Fear & Greed, CoinGecko, and Messari data. Uses contrarian logic (extreme greed = sell, fear = buy). 3. RiskGuardianAgent: A veto-only agent that blocks trades when PRISM risk scores hit β₯75, preventing over-leveraging. 4. MeanReversionAgent: Uses statistical deviation to identify overbought/oversold levels. Best in ranging markets. The Feedback Loop (Every 5 Mins) Data & Regime: Pulls live data (PRISM, Kraken, CoinGecko, Alternative.me) and classifies the market as RANGING or TRENDING. Voting: Agents independently vote (BUY/SELL/HOLD). Final direction is a weighted majority. Execution & Risk: Trades execute via Kraken only if all risk and compliance checks pass. Learning: Upon position close, PnL updates weights. Profitable agents gain influence (up to 2.0); losing agents drop (down to 0.7). Weight Evolution Timeline 0β8 Hours: Equal influence shifts to early divergence (weights 0.75β1.35). Small PnL fluctuations. 24β72 Hours: A dominant strategy solidifies (winning agent hits 2.0+ weight). Stable, trend-aligned profits emerge. Key Features & Readiness Transparent & Adaptive: No black-box AI; fully auditable decisions that adapt to current market regimes. Production Ready: 6 live APIs, 8 Flask endpoints, a professional dashboard, and tested dry-run uptime. The Win Condition NEXUS is ready for live capital deployment when one agent hits a 2.0+ weight, exceeds 55% accuracy, and yields a positive cumulative PnL. π
12 Apr 2026