Selected Work
Three engines,
in production.
Each one takes raw signal — market data, planetary positions, a scanned bill — and turns it into something a person can act on.
A
AAR · Autonomous Analysis & Recognition
Algorithmic trading · 2026
A production trading system of around 135 Python modules that watches 8 markets across 4 timeframes and trades them using the ICT / Smart Money Concepts method through more than 35 dedicated analysis engines.
Python 3.11Multi-threaded serverOANDA & MetaTrader 5Flask dashboardOpenAI GPT (vision)Telegram alertsDockerHetzner VPS · Cloudflare Tunnel
- Headless server runs a 5-minute scan loop, an entry watcher, a trailing-stop manager and the Telegram bot together, with a health-checked dashboard.
- A read-only backtester replays six years of history through the live engines, so the strategy I test is exactly the strategy that trades.
- Fixed 1% risk per trade, compounding — the system stays flat most of the time and only takes its highest-conviction setups: 1,138 trades in six years, not thousands.
- Validated walk-forward across all 8 instruments over six years — every instrument and every year profitable, so the edge isn't one lucky market.
Private system — happy to demo and walk through the live architecture on request.
O
Oracle Vedic Astro · Web Platform
Vedic astrology SaaS · 2026
A Vedic astrology platform with 187 computational engines that cross-check every result across 17 classical scriptures: full birth charts, life-area verdicts, timelines, kundli matching, remedies, a karmic reading, and a chat grounded in the user's own chart.
Python 3.12FastAPI · GunicornpyswissephOpenAI GPT (narration)SQLiteReportLab (PDF)Docker · NginxPaddle & Play billing
- Core design: engines compute, GPT only narrates. The AI can never invent a result, and every claim cites the planet, house and scripture behind it.
- Each of the 17 scriptures is its own rule module, so the engines cross-verify each other.
- Validated at 93.5% event-detection accuracy across 157 charts with known life events.
Oracle Vedic Astro on the web at oraclevedicastro.com.
O
Oracle Vedic Astro · Android App
Native app, Flutter · 2026
A real native app built in Flutter and published on the Google Play Store — one codebase, ready to extend to iOS and web.
Flutter · DartThin client over FastAPI backendSession-cookie authGoogle Play billing
- Calls the same backend that powers the website, so there is one source of truth and the app stays light.
- Every screen mapped to a backend endpoint, reusing the website's proven request and response wiring.
- Ported a detailed web design into Flutter widgets, screen by screen, with a custom theme system.
The Oracle Vedic Astro app on Google Play, built in Flutter.
P
PDF2Tally
Accounting automation · 2026
Takes a PDF purchase bill, or even a WhatsApp photo of one, reads it, and posts a complete purchase voucher straight into TallyPrime — creating any missing ledgers and stock items along the way. Built for my father, a practising accountant, and tested on his real vendor bills.
Pythonpdfplumber · PyPDF2OpenAI + Claude extractionTally HTTP / XML APITkintercustom Tally TDL
- Reads the bill, uses AI to pull out party, GSTIN, invoice number, line items with HSN, tax and totals, then posts it to Tally as a proper voucher in the Day Book.
- Checks Tally first and creates only the ledgers and stock items that are missing, so nothing is duplicated.
- Uses either OpenAI or Claude, so extraction keeps working across many vendor invoice layouts, including photographed bills.
- A custom TDL script adds an "Import PDF" button inside Tally, so it runs from one click in the software he already uses.
The PDF2Tally app, and the custom "Import PDF" button it adds inside TallyPrime.
A private tool I built for my father — in daily use in his accounting work.
