I am Founder and Principal Researcher at Capital Markets AI, building agentic AI platforms and publishing original research at the intersection of institutional risk and frontier AI. My career spans 20+ years in capital markets — Global Head of Technology Presales at Calypso Technologies ($50M ARR, 10 Solutions Architects across NA/EMEA/APAC), Director, Product Engineering at EZOPS (cloud-native AI reconciliation, 25-member engineering team, ISO-27001 attained in 3 months), and a recently completed Principal Software Consultant engagement at Fidelity Investments. Three SSRN working papers, an open Treasury PoC pairing Mandelbrot's MMAR with LLM-emitted regime signals, and three open-source projects on the Anthropic Claude API. I hold certificates in Quantitative Finance (Paul Wilmott) and Risk Management (Nassim Taleb) and hold the Claude Certified Architect – Foundations credential (Anthropic, 2026).
My core interest is the intersection where institutional capital markets domain depth meets frontier AI capability. I believe fractal risk intelligence — combining Mandelbrot's MMAR with LLM orchestration via Model Context Protocol — represents the most rigorous and verifiable reasoning domain available for training frontier AI models. I am building the case for this through open-source platforms, three SSRN working papers, and direct engagement with AI labs and institutional risk teams who are asking the same questions from opposite sides.
- 55-page combined paper: a seven-layer architecture positioning quantum compute at the tail-sampling layer for deep-tail VaR and Expected Shortfall, paired with a direct empirical measurement of the multifractal cascade’s loading-layer representability.
- Empirical result: bond dimension saturates the maximum 2^(K/2) across both lognormal and binomial multiplier laws — the cascade is not efficiently representable under MPS or balanced tree-tensor-network ansatzes, reopening the loading-layer question rather than closing it.
- Necessary but not sufficient for quantum advantage — no advantage is claimed; the deciding experiment is itself classical, and the two production-ready streams (multifractal mathematics + bounded LLM orchestration) stand regardless.
- 27-page original framework combining Mandelbrot's fractal mathematics (MMAR, α-stable distributions) with LLM agentic intelligence for market risk, counterparty credit risk (XVA, PFE, wrong-way risk), and liquidity risk.
- Original contributions: LLM + MMAR integrated system architecture, Hurst-based barbell sizing rules, fractal PFE for XVA repricing, Clayton copula wrong-way risk detection, three-tier platform design.
- Proposes AI firm partnership thesis: fractal risk as verifiable training signal for LLM mathematical reasoning, agentic tool orchestration, and calibrated uncertainty.
- 36-page institutional companion framework targeting CCPs, custodians, SIFIs, pension funds, and regulators — fractal precision for capital adequacy, sovereign architecture, and LLM-generated ICAAP/ILAAP supervisory narratives.
- Original contributions: fractal SIMM add-on factors by asset class, BCBS 248 intraday liquidity clustering model, joint CCR + liquidity tail modeling via Clayton copula.
- Includes Fractal Circuit Breaker regulatory proposal submitted to FSB, ESRB, FSOC, and BCBS — H-drift signals appeared 9–14 days before March 2020 dash-for-cash.
- End-to-end forward feasibility test on a $30M synthetic UST book (2Y/5Y/10Y, 10-trading-day window) pairing Mandelbrot's MMAR with an LLM-emitted six-dimension structured regime signal flowing through a deterministic, hand-recomputable mapping into Hurst (ΔH), volatility, and tail multipliers.
- Decomposed +38.5% E1→E3 lift in 10-day VaR_99 ($434K Gaussian → $602K Nexus-adjusted MMAR) into auditable channels: +19.8% pure MMAR self-similarity (text-blind, from t^H scaling) and +15.7% text-aware regime channel; designed Bar 2 forward discipline (single LLM, single pass, frozen corpus, score locked pre-window) to isolate calibration from comprehension.
- Full project documentation, executive summary, detailed findings, and methodological appendix with reproducibility from FRED-equivalent yields and four Python scripts; May 11, 2026 forward backtest publishes realized P&L against six pre-locked thresholds.
- Production-grade enterprise reconciliation platform (434 tests, 193 files) with Claude agentic AI — 7 agents for break investigation, broker onboarding, pattern analysis, config review, signoff briefings, natural language queries, and ML bootstrap training.
- Parallelized pipeline processing 100K+ records in under 300 seconds. Scored AWS Well-Architected 5/5, 12-Factor 12/12, Capital Markets Ops 10/10.
- Multi-cloud K8s deployment (AWS/GCP/Azure), REST API (20 endpoints), maker/checker signoff, DR (RPO 15min/RTO 30min), data residency enforcement (GDPR/MAS/SEC), SOC 2 Type II control mapping.
- Fully browser-based RAG application for capital markets professionals using the Anthropic API — single HTML file, no server, no install.
- 25-document knowledge base spanning Basel III/IV, XVA, SIMM, FRTB, SA-CCR, EMIR, ISDA netting, and trade reconciliation.
- Voice input (Web Speech API), retrieval trace panel, configurable project settings; showcased on LinkedIn in English, Chinese, and Hindi.
- End-to-end Excel automation toolkit using Claude AI — from concept, coding, testing to GitHub deployment.
- Reusable Python macro (openpyxl) and Excel VBA alternative for CUSIP LEFT(3) transformations with live formulas and scalable batch pipeline.
- Building agentic AI platforms and publishing original research bridging institutional capital markets risk and frontier AI.
- Three SSRN working papers (105 pages combined) — "A New Era for Capital Markets" (Abstract 6584378, 27 pp.), "A New Era for Institutional Finance" (Abstract 6615841, 23 pp.), and "Quantum-Augmented Risk Management for Capital Markets" (Abstract 6874958, 55 pp.).
- Active Treasury PoC V1.5 pairing Mandelbrot's MMAR with LLM-emitted six-dimension regime signals on a $30M synthetic UST book — decomposed +38.5% lift in 10-day VaR_99 ($434K Gaussian → $602K Nexus-adjusted MMAR) into auditable channels under Bar 2 forward discipline; May 11, 2026 forward backtest publishes realized P&L against six pre-locked thresholds.
- Three open-source projects on the Anthropic Claude API: a multi-agent Capital Markets Reconciliation Platform (434 tests, 193 files, 7 agents), CM·RAG (browser-based RAG over 25 capital markets regulatory documents), and CUSIP Transform Toolkit. Cross-model adversarial review across Anthropic, Google DeepMind, and xAI for all published research.
- Increased operational efficiency by 20% on Calypso implementation for collateral management, repo, and securities lending.
- Assisted Fidelity's onboarding team across product rollout, report generation, and clearing validation.
- Led Calypso implementation for collateral management, repo/lending, and interest/inflation swaps.
- Led full sales cycle from POC scoping through contract close with tier-one capital markets institutions, increasing ARR by $15M for AI-powered compliance and reconciliation platforms.
- Architected Java microservices, HTML5 interfaces, and TensorFlow-based ML algorithms, reducing data reconciliation time by 20% on 10M+ row datasets.
- Built and managed 25-member engineering team in Chennai, attaining ISO-27001 certification in three months.
- Owned full sales cycle from prospecting through close, selling $50M ARR in enterprise risk solutions to CME, HKMA, SGX, BOVESPA, Citigroup, HSBC, and other major institutions.
- Built and managed a global team of 10 Solutions Architects across North America, EMEA, and APAC.
- Architected enterprise risk products for market, credit, XVA (CVA, DVA, FVA), and liquidity risk with Basel III compliance.
- Manager/Architect, Global Delivery at Pinnacle Systems Inc.
- Manager/Architect, Global Delivery at IBM Global Services.