Solo product · Production
TradeBook
A full-feature trading journal with AI trade coaching, F&O Greeks analytics, psychology tracking, and dual-jurisdiction tax reports. Designed and built solo on Next.js 15.
What it does
A journal that gives feedback, not just storage.
TradeBook is a journal and analytics layer for active traders. Log trades, including F&O positions; the system pulls underlying prices, calculates Greeks, and stores entries with a psychology tag.
The product is built on a specific bet: the gap between a trader and consistent profitability is not market knowledge but pattern recognition over their own behavior. Most traders journal in spreadsheets that rot or in apps that store data without surfacing patterns. TradeBook flips that: an LLM-driven coaching layer reads every trade and surfaces patterns in weekly review.
The dual-jurisdiction tax module is the segmentation move. Indian traders who file ITR-3 and US traders who file Schedule D both get a single export at year-end. NRIs and global traders who file in both jurisdictions get the rare benefit of a single source of truth.
Features
The five layers that turn a journal into a coach.
Trade entry with auto-pulled context
Logging an F&O position auto-pulls the underlying, calculates Greeks (delta, gamma, theta, vega), and stores the entry. The user adds a 1 to 2 sentence rationale at entry; once the trade closes, an exit note plus a psychology tag.
AI trade coaching
Claude reviews each closed trade — entry rationale, position sizing, exit, psychology — and writes a coaching response. Over time the engine surfaces patterns: cutting winners early, holding losers too long, sizing inconsistently. The kind of feedback you'd get from a senior trader, except on every single trade.
F&O Greeks analytics
Real-time Greeks calculation for active positions. Greeks aggregate at the portfolio level so the trader can see total delta exposure, theta decay per day, and vega risk across the open book.
Psychology tracking
Each trade gets a tag — discipline, FOMO, revenge, planned. Over weeks, the journal surfaces the correlation between psychology tags and outcomes. A trader who tags 30% of trades 'FOMO' and underperforms benchmarks on those exact trades has actionable evidence.
Dual-jurisdiction tax export
End of year, the tax module exports either an ITR-3 P&L statement for India or a Schedule D summary for US filings. Same trade entries, two jurisdictions, no manual reconciliation. This is the market-segmentation move that makes the product valuable to NRIs filing in both countries.
Try it live
TradeBook is deployed and operational.
The full app is live on Vercel. Click through the journal, coaching panel, Greeks dashboard, and tax export.
Open tradebook-sepia.vercel.appArchitecture
Server Components for data, Server Actions for mutations.
Stack: Next.js 15 with the App Router. Edge functions for Greeks calculation (low-latency math on the underlying spot price). Server Components for the dashboard data-heavy views. Server Actions for trade-logging mutations.
AI layer: Claude API wraps the coaching engine. Trade context (entry rationale, size, exit, psychology) is structured into a consistent prompt so coaching responses are comparable over time. The coaching runs asynchronously after trade close — the user sees feedback a few seconds later, not blocking the close action.
Tax engine: Jurisdiction-specific logic for ITR-3 (India) and Schedule D (US), driven by the same trade entries. The engine handles short-term vs long-term classification, F&O P&L aggregation under business income (India), and lot-by-lot matching for capital gains (US).
Data: Postgres for trades, edge-cached for read-heavy paths. Authentication via a session cookie with edge-validated JWT.
Why this matters
For mid-market AI implementations.
TradeBook is a different kind of demonstration than Ask Data. It shows that I can take a complex, segmented domain (trading, with regulatory variation across jurisdictions), specify the product deeply, and ship a multi-layered application end-to-end.
The same pattern transfers cleanly to mid-market clients whose problems are not generic. A logistics company with a bespoke routing problem. A healthcare admin firm with two payer formats. A SaaS company with a billing model that does not fit Stripe out-of-the-box. These are the engagements where senior BA discipline plus AI implementation produces outsized results, because the off-the-shelf vendors cannot serve them.
If your operational problem has that shape — specific, multi-layered, with no obvious vendor — the discovery call is the next step.
This is the kind of full-stack AI product I ship for clients.
If you want something like this for your operations or your customer-facing product, the discovery call is the next step.