Kyle Weinstein

An equity research pipeline where every factual claim traces to a specific, dated, tier-labeled primary source. Information sourcing and citation verification are enforced deterministically by code wherever possible to minimize reliance on LLMs, reducing both cost and hallucination risk.

• Seven-tier source hierarchy from SEC filings down to unaudited public posts, with market data marked as a point-in-time price fetched at the time of the research.
• Claims are extracted by regex, not by a model. The citation grammar is generated from one module the pipeline and the prompt both import, so they cannot drift apart — the code raises at import if they do.
• Mechanical tie-out with a digit guard. A quoted figure is located in the source by exact search; a one-digit change leaves a ~97%-similar string, so similarity alone would confirm a fabrication. Comparing digits does not.
• The audit separates proven from judged instead of averaging them into one reassuring number, and deep-links each proven claim to the exact sentence in the filing.

An evidence-first equity research workflow that turns SEC filings into thesis-driven reports and a reviewable monitoring queue. The factual report is verbatim source evidence—not AI paraphrase—so each published claim can be traced to a dated, accession-numbered filing.

• The customer-tested workflow is preserved end to end: watchlist → thesis-driven report → EDGAR sync → review queue → evidence drawer → disposition → audit export. It operationalizes the research standard, "tie it out or it doesn't go in."
• Evidence is quote-first. Filing sections are located deterministically; a structured-output model selects relevant verbatim excerpts and classifies them to monitoring dimensions without rewriting the underlying source text.
• Verification is mechanical where it can be. Normalized substring matching and a digit guard reject altered quotations, while explicitly labeled derived analysis receives a separate semantic check. Failures are quarantined for review, never silently corrected into the report.
• The workflow is durable and schedulable. PostgreSQL-backed jobs, reconciliation, and server-owned progress survive navigation and refresh; scheduled EDGAR checks compare accession metadata first, making a no-new-filing result a zero-model-cost no-op.

An interactive laboratory for designing, stress-checking, pairing, and rebalancing long-only portfolios from adjusted daily-close history. It keeps the analytical model transparent while making dense portfolios fast enough to explore directly in the browser.

• One diversification map scales continuously. A clustered Canvas explorer moves from portfolio overview to exact pair values through semantic zoom, search, pan, and keyboard navigation instead of switching between portfolio-size-specific designs.
• Large analyses stay off the interface thread. Correlation, classification, clustering, and constrained optimization run in a Web Worker over packed typed arrays, keeping draft editing responsive as holdings grow.
• Pairwise work is reduced to O(N²T). Return series are aligned and standardized once, the upper correlation triangle is stored once, and dependent tools reuse the same matrix rather than rebuilding statistics for every cell.
• Drafts and analyzed results are deliberately separate. Expensive work runs only when requested, prior results remain visible while inputs change, and stale requests cannot overwrite a newer portfolio snapshot.

A browser-based educational game designed to master market-making fundamentals and ace quant trading interviews.

Background: Why I Built This

I was preparing for interviews at quantitative trading firms like Jane Street, Optiver, and Tower Research—the kind of places where they ask you to think like a market maker in real time.

The problem? Most market-making resources are either:

• Too theoretical (textbooks with no interactivity)
• Too narrow (focus only on one concept)
• Not interview-focused (don't match the actual questions asked)

So I built this game to consolidate everything I was learning into one place where I could:

• Learn systematically - from order book basics to strategic positioning
• Practice repeatedly - with infinite freshly-generated problems
• Simulate real scenarios - competing against AI, managing risk under pressure
• Get feedback immediately - understanding why each answer was right or wrong

The result is a game that teaches the actual skills tested in quant trading interviews while being genuinely fun to play.

What You'll Learn

Order Book Mechanics

• Reading and interpreting order books
• Calculating execution prices and slippage
• Understanding liquidity and market depth
• Analyzing order flow patterns

Market Making Strategy

• Capturing the spread without getting "picked off"
• Managing adverse selection risk
• Positioning orders strategically
• Balancing profitability vs. fill rates

Quantitative Thinking

• Price impact calculations
• Risk/reward tradeoffs
• Position management
• Statistical pattern recognition

Interview-Specific Skills

• Quick mental math under time pressure
• Clear reasoning about market microstructure
• Strategic decision-making with incomplete information
• Communication of trading ideas