🤓 Latest Submissions
Project Doomsday
Project Doomsday is an quality financial stress-testing engine that simulates how Black Swan events destroy company valuations in real-time. Enter any listed stock ticker — US or Indian; the system deploys a coordinated swarm of 5 AI agents powered by Google Gemini 2.0 Flash via Google AI Studio: 1. Intelligence Analyst: Executes 6 parallel search vectors (geopolitical, supply chain, financial, regulatory, technology, macro) using Tavily real-time search, then synthesizes findings into 6 specific, geolocated risk scenarios. 2. Bear Advocate: Prosecutes each risk with worst-case arguments, historical precedents, and quantified impact estimates. 3. Bull Advocate: Challenges evidence, presents mitigating factors, argues market pricing efficiency. 4. Fracture Judge: Delivers calibrated verdicts (8+ = catastrophic, 6-7 = material, <4 = dismissed) with final severity and probability scores at low temperature for high-conviction decisions. 5. Contagion Modeler: Traces how primary shocks cascade into second, third, and fourth-order effects through credit linkages, supplier networks, and investor psychology. Validated risks are plotted on a global fracture map with curved convergence lines showing geographic vulnerability concentration toward company HQ. The valuation engine uses 5-path context-aware routing — P/BV for banks (debt is their product), EV/Revenue for high-growth (negative FCF breaks DCF), normalized EBITDA for cyclicals (spot earnings mislead at peaks), full DCF only for mature profitable companies, and capped revenue multiples for loss-makers. Base fair value is market-anchored, then stressed through revenue haircuts, margin compression, multiple de-rating, and WACC premium injection. A chaos slider (0-100%) lets users sweep from mild headwinds to full systemic doomsday instantly without re-running the AI pipeline. Every calculation is transparent — full audit trail showing routing logic, formulas with numbers plugged in, and stress decomposition.
19 May 2026
Project Veritas
Project Veritas is a professional-grade intelligence engine designed to automate the high-stakes workflow of Private Equity due diligence. Traditional due diligence takes weeks of manual data gathering and subjective debate; Veritas reduces this to sub-minute cycles without sacrificing rigor. The system operates in four specialized phases: High-Fidelity Extraction: Using custom Python tools, it pulls TTM (Trailing Twelve Months) financials via yfinance and SEC EDGAR, ensuring a "ground truth" data layer. Peer-Set Benchmarking: It dynamically generates CapIQ-style peer groups to validate valuation multiples and industry-standard margin parity. Institutional RAG: A ChromaDB layer, indexed with 40+ institutional finance textbooks, ensures all mathematical derivations (DCF, Multiples, QofE) follow strict industry methodologies. Multi-Agent IC Debate: The core innovation is an adversarial debate between a "Deal Champion" (Bull) and a "Risk Partner" (Bear). They challenge each other’s assumptions on forensic red flags and quality of earnings, arriving at a final, defensible Investment Committee (IC) verdict. By offloading the "heavy lifting" to a multi-agent swarm, Veritas allows analysts to focus on high-level strategy while maintaining an audit trail of every decision.
10 May 2026
