Projects

Operator-first market-state instrument estimating the multiscale flow-to-price transfer operator under hard admissibility constraints, decomposing realized covariance into endogenous fragility and exogenous shocks.

Information-geometric market metrology system computing Fisher-Rao manifold curvature to identify which directions of market belief are fragile, robust, and likely to amplify shocks.

Reconstructs the financial system's latent intermediation-constraint state from cross-market law-of-one-price wedges, converting CIP, Treasury basis, and swap spreads into audit-traceable constraint signals.

Measures the minimum deformation required to move implied futures into a jointly admissible world, surfacing dual tension, analog states, and failure artifacts before favorable outputs.

Deterministic public-rule activation instrument that measures where installed capital becomes mechanically compelled to act, using only public disclosures, codified mandates, and benchmark definitions. The first domain is the leveraged and inverse ETF daily-reset rule, replayed against SEC EDGAR filings with SHA-256 provenance, ACCEPT, REROUTE, KILL, and UNIDENTIFIED memos, and honest reporting of a pre-registered gate that does not clear.

A measure-valued reconstruction of still-outstanding capital, indexed by entry basis, age, carry, cohort, and sign, evolving under a birth, ageing, and death process. Built on a Bitcoin UTXO oracle over 189 million transaction outputs, it maps the disclosure-fidelity boundary at which monthly NPORT-P cadence becomes insufficient to identify the vintaged ledger.

An ex-ante measure of how much publicly scheduled institutional trading mass from distinct calendar families is time-aligned to the same closing window. Across a fifteen-ETF panel it isolates a non-spurious overlap term whose directional signature favors information aggregation over transient pressure, with four pre-registered directional hypotheses rejected and committed as artifacts.

A conserved, dated inventory of facts that are already legally true but not yet public, treating the cross-statute schedule of forced future disclosure as a measurable market-state object. An exact timing layer is computed deterministically from regulatory rules while magnitudes are reported as partial-identification bounds, validated on EDGAR filings with integrity gates and a Schedule 13D natural experiment.

An IC-grade deal intelligence suite, deterministic scenario simulation + corporate-graph cascade modeling + SEC/XBRL pipelines: delivering audit-traceable, memo-ready decision outputs.

Axiom Atlas, a 3-page desk-grade finance platform unifying a Market OS (live macro, filings, IPO intelligence), a Stock Lab (OHLC charting + probabilistic price forecasts), and a real-time News Wire, plus a Gemini 2.5 AI Product Guide that answers any query, parses uploaded PDFs, preserves multi-thread chat state, and routes users to the right tools and next actions with traceable outputs.

Governance-first agentic investing MVP with deterministic risk controls and hash-chained audit trails that make agent decisions reviewable, reproducible, and safe to iterate.

This platform stress-tests one deal across multiple plausible futures so you can see what breaks and what survives. It is an investment-committee operating layer that converts scenario analytics into decision-grade risk, return, and mitigation evidence for live transactions.

Ingested SEC XBRL Company Facts, normalized US-GAAP tags, standardized statements, computed TTM KPIs and trading multiples with exportable artifacts.

Walk-forward research pipeline with standardized KPIs (CAGR, Sharpe, Sortino, MaxDD, turnover) and reproducible runs designed for auditability.

Python-based valuation engine for Monte Carlo DCF + Reverse-DCF back-solve to EV/IPO targets; exportable artifacts + memo-style outputs.

Sources & Uses, multi-tranche debt, covenant-style checks, and downside stress tests to evaluate returns and fragility.

A structured workflow for comps, narrative, and model artifacts—built for speed, consistency, and clean outputs.

Character-level BiLSTM + attention to predict English lexical stress patterns using CMUdict.

Driver-based 3-statement model + 10-year unlevered DCF; downside and accretion/dilution + synergy analysis to set a bid ceiling and diligence priorities. Deal materials (Model + IC memo).