Sandeep, drawn as a cartoon with curly hair, glasses and earphones

trades as Gravity India

I build SaaS products, trading infrastructure, and desktop tools.

Most of what I ship begins as a bottleneck on my own trading desk — custom Canvas 2D charts, Level-3 orderbook ingestion, NSE F&O rotation scanners, macOS Swift widgets, and public web products. I build it from bare metal, run it every session, and keep sanding it down until it earns its place.

13+Shipped builds & tools
62Custom chart indicators
223NSE F&O stocks scanned
4Markets · NQ, GC, Nifty, F&O
drag the stickers,
they bounce back

Work & Projects

13 projects
2026
  • The charts I trade from every session. Unifies Nasdaq & Gold futures (Rithmic / Probe MBO feeds), Nifty 50 index, Nifty futures, and the full NSE F&O cash-equity universe (Dhan) onto one custom Canvas 2D engine with plain ES modules and Node.js 24+.

    • 62 built-in indicators: session & visible volume profiles, MA Ribbon, key levels, CVD by trade size, flow dominance, depth pressure studies, and macro overlays.
    • Full drawing suite: pitchforks, Fibonacci/Gann, regression channels, bar pattern replay, and locked/templated rectangles with corner & midpoint handles.
    • Built-in Agents worker: generates validated declarative tools and bounded rolling expressions in an isolated flow without arbitrary JS execution.
    Canvas 2DES ModulesNode.js 24+Rithmic & Dhan62 Indicators
  • One clean place to check when a movie lands on Indian streaming platforms. Every date goes up only after a person checks the primary source — eliminating speculative rumor calendars — backed by an Astro storefront, Python HTTP backend, and SQLite WAL storage.

    • Source-checked calendar: human review gate ensures announced OTT dates stay trustworthy.
    • Launch & waitlist pipeline: SQLite WAL persistence with an automated approval queue worker.
    • Zero-downtime edge hosting: served behind Cloudflare with instant static page delivery.
    AstroPythonSQLite WALCloudflare ottflix.in ↗
  • The command center for the Gravity trading desk. Brings live equity, open positions, fills, risk sizing rules, and post-trade exit reviews into one view alongside a full-universe NSE F&O scanner tracking intraday setups and monthly structure bands.

    • 223 NSE F&O instruments: scanned continuously from live Dhan streams over Server-Sent Events (SSE).
    • Dual paper & live books: enforces pre-trade risk limits and logs every entry, partial, and exit for review.
    • FastAPI + SQLite candle store: low-latency local bar persistence for instant setup scoring.
    FastAPISQLiteDhan SSE223 NSE F&ORisk Rules
  • Money rotates continuously across pockets of the market. Gravity Rotation scores where sector leadership is accelerating or fading each session, while serving as a shared internal quote fanout so downstream dashboards never open duplicate broker sockets.

    • Sector leaderboard: ranks top performing and lagging NSE sectors dynamically through the session.
    • Single-socket quote bridge: solves broker connection limits by multiplexing live Dhan WebSocket ticks cleanly.
    • Opening runner feed: pairs with early-session logic to spotlight morning leadership right after the open.
    PythonDhan WebSocketSector RankingQuote Fanout
  • Three desktop-level macOS widgets compiled straight with Apple swiftc — no Xcode bloat, no Electron RAM tax. They sit pinned above the wallpaper across all Spaces for zero-friction desk awareness.

    • Gravity Deck: unified card grid for live equity & P&L, IST/ET NSE & CME countdowns, 09:17 runner state, top-5 sector rotation, VWAP/levels, and feed latency.
    • Kira Fleet Widget: polls systemd units, Docker containers, timers, RAM, and disk every 20s with explicit staleness badges.
    • Quote Widget: borderless click-through desktop card polling NQ, GC, and NIFTY prices every 15s.
    SwiftAppKitPython Collectorssystemd
  • Closed candles hide who actually defended a price level. This market-data pipeline captures continuous Rithmic/Databento Market-By-Order (MBO) queue events for NQ and Gold futures into ClickHouse, paired with an automated absorption & S6 detector.

    • ClickHouse ingestion: streams raw ticks and L3 queue events into market.trades with daily zstd archival and verified backups.
    • Absorption & S6 detector: flags resting limit reloads absorbing aggressive sweeps before price reverses.
    • Shadow evaluation & alerts: logs structured shadow JSONL outcomes and pushes instant Telegram notifications.
    ClickHouseRithmic MBODatabentoPythonzstd
  • An AI-assisted futures backtester for NQ and GC built so nothing is silently assumed. You describe a strategy; the model proposes quote-backed rules that require explicit approval, locks a SHA-hashed spec, previews trades on the chart, and runs the exact same engine on Databento DBN ticks.

    • Native Databento DBN reader: ingests .dbn / .dbn.zst (MBO, trades, TBBO, OHLCV), merges 18:00–17:00 ET CME sessions, and resolves contract roll days.
    • True intra-bar tick replay: replays exchange-timestamped prints whenever both stop and target hit inside one bar.
    • Hash-locked execution: refuses backtests unless the spec hash matches the user-approved preview version.
    PythonDockerDatabento DBNDeepSeek GitHub repo ↗
  • A public web product translating internal screening workflows into a fast, zero-bloat web surface at openscreener.xyz, served via Caddy and Cloudflare Tunnel.

    • Edge-routed architecture: static delivery behind Caddy reverse proxy and Cloudflare Tunnel.
    • Clean responsive interface: built with zero heavy client bundles for instant load times.
    WebCaddyCloudflare Tunnel openscreener.xyz ↗
  • Two autonomous AI infrastructure tools built to inspect servers and evaluate trade setups from anywhere without living inside a terminal.

    • AGY Telegram Plugin: systemd daemon with 6-digit OTP pairing, live typing feedback, /effort & /mode controls, chunked diffs, and 0.1s shell bypass.
    • Alpha Council: orchestrates 8 specialist bots (MacroHawk, WhaleTracker, GammaOptions, ChartStructure, SocialSentinel, CatalystHunter, ValueAuditor, RiskOfficer) to flag bull/bear traps and thesis vetoes.
    PythonsystemdTelegram Bot APIMulti-AgentGEX & Flow
  • Built two focused tools for studying writing craft and publishing cleanly: X Library keeps a private SQLite archive of standout posts scored by follower-normalized outlier ratio; Orbit maps out queues, signals, and voice-matched drafting in Next.js.

    • Outlier scoring math: divides weighted engagement by the median for the author’s follower band so small-account breakouts rank fairly.
    • Zero-telemetry local storage: single Python process + SQLite FTS for instant hook & format search.
    Next.jsPythonSQLite FTSTypeScript
2025
  • Translated custom Pine Script supply & demand zone logic across the entire Nifty F&O universe into a real-time FastAPI + React dashboard showing only stocks touching or breaking upper/lower zones today.

    • Multi-timeframe zone engine: tracks daily, weekly, and monthly touch, break_up, and break_down events via live FYERS ticks.
    • TradingView & Discord export: one-click symbol-list copy per timeframe plus segmented Discord webhook alerts.
    FastAPIReactViteFYERSDiscord Webhooks
  • A Next.js 15 EOD signal dashboard designed for honest post-market evaluation — computing entry returns from the shifted next-day open price rather than stale close prints, paired with stock detail charts and trade journal panels.

    • Bias-free EOD returns: anchors entry price to the next calendar session’s opening candle.
    • Read-only broker sync: optional Supabase persistence + FYERS EOD candle synchronization.
    Next.js 15TypeScriptTailwindRechartsSupabase
  • Public repositories maintained under @Sandeep-09 covering historical market dataset builders, equity screeners, Telegram automation bots, and early utility builds.

    • Market utilities: Historify, Screener, and quant-spotlight-app for data preparation and screening.
    • Automation & apps: Telegram-bot and early engineering repos going back to Metron.
    GitHubPythonTypeScript @Sandeep-09 ↗
filter by craft or
tap a row for specs

Trading as Gravity

NQ · GC · Nifty · F&O

Gravity is the name on my trading desk. I trade with pre-defined rules, size strictly by risk, and journal every session. Most of the software above exists because the desk needed faster tape visibility or cleaner risk control.

Nasdaq futures (NQ) Gold futures (GC) Nifty 50 223 NSE F&O stocks

Order flow on the L3 DOM

Candles tell me what already happened. The Level-3 Market-By-Order book shows it happening: every resting limit order in the queue, who is reloading into market sells, who is pulling liquidity, and which side absorbs the flow.

Live simulated queue: watch resting bids & asks reload, pull, and hit the tape in real time.
L3 market by order DOM ladder Time & sales
01 · TAPE FIRST L3 Queue & Absorption

Read resting depth reloads and swept offers directly off Rithmic & Databento MBO streams rather than lagging overlays.

02 · TWO SESSIONS India & US Rhythm

NSE cash F&O rotation & Nifty futures from 09:15–15:30 IST, followed by CME Nasdaq & Gold futures in NY hours.

03 · CODE-BACKED Replay Before Size

Every discretionary idea is logged in shadow mode or tick-replayed on Databento history before it earns live risk.

Gravity =
my trading name

Stack & Infrastructure

Zero-bloat tools
Languages
JavaScript (ES Modules)TypeScriptPythonSwiftSQL
Frontend & UI
Canvas 2DNext.js 15AstroReactTailwind CSS
Backend & Data
Node.js 24+FastAPIClickHouseSQLite (WAL)Supabase
Market Feeds
Rithmic MBODatabento DBNDhanFYERS
Infra & Edge
Linux VPS (systemd)DockerCloudflare Pages & TunnelsCaddy
Workflow & AI
Claude CodeCodexDeepSeek / OpenRouterTelegram Bot API

Bookshelf

Tap a spine

One Piece, Vol. 103 by Eiichiro OdaLong-arc worldbuilding, patience, and building a crew that stays sharp under pressure.

manga next to
money, on purpose

Hello !

  • I'm Sandeep. I build SaaS products from the first rough sketch to live deployments, data models, and edge hosting.
  • I trade the markets as Gravity and most of my software began as a missing tool on my own trading desk.
  • I bias toward lean, fast systems — raw Canvas 2D over heavy chart wrappers, single-binary Swift widgets over Electron, SQLite & ClickHouse over unnecessary cloud sprawl.
  • Right now I'm refining Kira Charts, GravityAlpha, and OTTFLIX while wiring autonomous AI research tools and tick-replay validation into daily trading workflows.
Explore shipped products & code: