Skip to main content

All Systems in Action

32 production-ready systems — what each one does, how it's built, and the stack behind it. Walkthroughs available on request.

Showing 8 of 32
Trading SystemsFlagship

Enterprise Algo Execution Platform

Multi-broker execution engine with a unified control surface

A production trading platform that routes orders across two brokers from one Next.js control surface. Every signal clears a five-check risk gate before it can reach an exchange — kill switch, circuit breaker, liquidity, capital adequacy, daily loss — and each refusal is logged with the reason, so a trade that never happened can still be explained.

Pre-trade checks
5
Broker adapters
2
Strategy engines
2 live · 6 configured
  • Single unified order API over Zerodha and Flattrade
  • Five-stage risk gate between every signal and the exchange
  • Every refusal logged with its reason code, not silently dropped
  • Next.js control surface reporting every stage of the pipeline
  • Telegram alerts for signals, fills, stop-loss and target hits
PythonAsyncIONext.jsWebSocketTelegram
Enterprise AIFlagship

Ethereal Website Audit Engine

Three AI agents audit a site in parallel and merge into one report

Point it at any URL and three specialised agents review it simultaneously — one for UX, one for marketing, one for business model. A compressed snapshot layer distils the rendered page before it reaches the model, and the compression ratio it achieves is logged on every run.

Parallel agents
3
Model calls
3 concurrent, capped
Key pool
Rotates on quota
  • Three specialised agents (UX, marketing, business) running in parallel
  • Structured snapshot replaces the raw page dump before any model call
  • Playwright renders JavaScript-heavy pages before analysis
  • Findings merged and de-conflicted into a single prioritised report
  • Streamlit interface with exportable audit output
PythonPlaywrightGemini 2.5 FlashStreamlitMulti-Agent AI
Enterprise AIFlagship

AI Options Strategy Copilot

Plain-English intent in, precise multi-leg structure out

Type "I'm bullish but want downside protection" and it selects the matching options structure, computes the strikes, and stages execution across every linked account, unwinding any partly-filled basket. The split matters: an LLM classifies intent only, while every strike, spread width, and lot size is computed by deterministic Python — so the model can never invent a number that reaches your broker.

Strategy templates
24
Strike calculation
Deterministic
Live P&L push
2s (15s off-hours)
  • Natural-language intent classification mapped to defined strategy templates
  • All strike and payoff maths computed deterministically, never model-generated
  • Partial-fill unwind — a basket that cannot complete is reversed leg by leg
  • Payoff preview and margin check before any order is placed
  • Live position tracking over a persistent WebSocket feed
PythonGoogle GeminiZerodha KiteConnectWebSocketFastAPI
Trading SystemsFlagship

RSI Scalping React Platform

Full-stack scalping system with autonomous exit management

Evaluates RSI two-candle body reversals at each one-minute candle close and manages the whole trade lifecycle without intervention. Once filled, the entry is split into three separately-tracked target legs at 1.5R, 2.0R and 2.5R, and open legs are rehydrated from the trade log on restart so a crash never orphans a position.

Scan cadence
60s close · 5s LTP
Profit tiers
3, independent
Exit handling
Autonomous
  • RSI(4) two-candle body reversal evaluated at each 1-minute candle close
  • Each entry split into three target legs (1.5R / 2.0R / 2.5R), tracked separately
  • Autonomous exit management — no manual intervention after entry
  • Engine heartbeat with a 120-second stall alarm, surfaced in the dashboard
  • Typed FastAPI backend with full order audit trail
PythonFastAPINext.jsTypeScriptZerodha KiteConnect
Trading SystemsFlagship

GSheets Supertrend Signal Bot

Automated Supertrend execution with a human-readable audit trail

A Supertrend options bot built around an explicit state machine, so the system is always in exactly one known state and cannot double-enter or orphan a position on restart. Every transition writes to Google Sheets and CSV, giving a complete trade record you can read without a database client.

Execution model
State machine
Audit trail
CSV + JSON status
Concurrency guard
Exclusive lock
  • Explicit state machine preventing double entries and orphaned positions
  • Every lifecycle event written to a CSV audit trail with source and severity
  • Automatic signal capture — no missed alerts from manual monitoring
  • Full CSV audit trail alongside the live sheet
  • Exclusive process lock — a second bot cannot start against the same account
PythonGoogle Sheets APIZerodha KiteConnectState MachineCSV Audit
Enterprise AIFlagship

ServiceNow Change Validator

AI middleware that reviews Change Requests before a human does

Sits between ServiceNow and your change approval board, automatically assessing every Change Request for compliance gaps, risk level, and missing detail. A PII scrubbing layer strips identifying data before anything reaches the model, so the integration stays within enterprise data-handling policy.

Review time
30 min → ~6 min
Data handling
PII-scrubbed
Integration
Native ServiceNow API
  • Automated compliance, risk, and completeness assessment per request
  • PII scrubbing layer applied before any model call
  • Cuts a ~30-minute manual review to roughly 6 minutes
  • Structured findings written straight back to the ServiceNow record
  • Designed to satisfy enterprise data-handling review
PythonGoogle GeminiServiceNow APIFlask
Trading Systems

Legato Strategy Execution Dashboard

See a multi-leg strategy as one position, not three orders

Brokers show you three separate legs; Legato shows the combined strategy. It aggregates legs into a single position with unified P&L and exits them in a deliberate order — short legs first to release margin, hedges second. A complete mock adapter stack sits behind the same interfaces as the live broker, so the whole product demonstrates end to end without an account.

Price feed
KiteTicker + 1s reconcile
Demo mode
No credentials needed
Licensing
Hardware-locked
  • Multi-leg positions aggregated into one combined P&L view
  • Sequenced exit — short legs first to release margin, hedges second
  • Sub-second MTM pushed over WebSocket, with a 1s REST reconcile
  • Strategy-level trailing stop-loss across the whole structure
  • Full mock adapter stack — runs end to end with no broker account
PythonFastAPIWebSocketZerodha KiteConnectPyInstaller
Trading Systems

Fair Value Gap Options Trader

Dual-methodology FVG detection across three indices at once

Implements two independent Fair Value Gap detection methods — classic ICT and the LuxAlgo variant — and lets you trade either immediate entry or gap-touch confirmation. It rescans NIFTY, BANKNIFTY and SENSEX on a ten-second cycle, detecting only on closed bars so a signal can never repaint, with every signal and outcome persisted to PostgreSQL for later analysis.

Detection methods
2 (ICT + LuxAlgo)
Scan cycle
10s, 3 indices
Signal storage
PostgreSQL
  • Two independent FVG detection methods: ICT and LuxAlgo
  • Detection runs only on closed bars — signals never repaint
  • Three indices rescanned every 10-second cycle
  • Per-instrument stop-loss, target, trailing SL and premium band
  • Every signal and outcome persisted to PostgreSQL for review
PythonPostgreSQLStreamlitZerodha KiteConnectFVG Logic