A trading platform built around a professional trader's experience
A platform built around a trader's many years of experience: its own trading core, recording and replay of the trading day, a terminal with analytics and AI tools. It automates routine work and strengthens decisions, runs live and is updated regularly.
- Client
- NZT
- Engagement
- Custom development and support
- Industry
- Algorithmic trading on the Moscow Exchange
- Timeline
- since summer 2026, development and support continue
- Scope
- Platforms, Integrations, Bots
- Technologies
- Python, FastAPI, SQLAlchemy, PostgreSQL, Redis Streams, Celery, Next.js, TypeScript, Custom canvas chart, Docker, gRPC
[02] Task
What needed to change
To carry the approach of a professional trader with many years of experience into a system used by the trader and the team of his investors. Routine work should run automatically, every decision of the system should be transparent and verifiable, and new ideas should be quick to test on data.
Scale 35 instruments at once, 5 to 6 million market events per trading day
What was there before us
- Watching dozens of instruments by hand
- Reviewing trades from memory and notes
- Testing ideas manually, with no quick way to run them on real days
Why nowThe trader needed a platform that grows with him: new ideas and tools are added without stopping the work.
[03] Solution
What we did
-
01
Our own trading core
An event-driven core makes decisions in fractions of a millisecond across dozens of instruments at once. Data intake, the engine, execution and accounting are separate services, and the database is not in the path of a decision.
-
02
Recording instruments and replaying the trading day
Selected instruments are recorded through the day. Any fragment can be replayed through the same code that trades (Polygon), so an idea is tested on a real day before it goes live.
-
03
Our own terminal with analytics
A chart with an order book heatmap and levels, a decision journal with a step-by-step review, rewind through the day and a Monitor with all instruments on one screen.
-
04
AI tools for analysis and for developing the system
AI helps analyse the recorded data, review decisions and speeds up work on the system. Trading decisions are made by the core under fixed rules, not by AI.
-
05
Investor portal and notifications
Investors see their account and receive Telegram notifications. Messages contain no amounts or percentages.
-
06
Off / shadow / live modes and a mobile version
In shadow mode the core makes decisions but sends no orders, so new logic is switched on safely. The Monitor, the chart and the operating mode are available from a phone.
What the solution consists of
- Platformsa web terminal with a mobile version, an investor portal, an API, the trading core with a decision journal, recording and replay of the trading day
- Integrationsthe broker API: market data streams, orders, operations; the exchange trading calendar; accounting reconciliation
- Botsa Telegram bot and client notifications
- Trading core
- Recording and replay of the trading day (Polygon)
- Terminal with a chart
- Decision journal
- Monitor
- AI tools for analysis and development
- Mobile version
- Investor portal
- Telegram bot and notifications
Stages and timing
-
summer 2026
Platform launch
A platform with workspaces and accounting, the first trading scenarios on a system of its own.
-
summer to autumn 2026
Terminal and trading core
The core and a terminal with analytics: a chart with the order book, the decision journal, the Monitor.
-
autumn 2026
Recording and replay, investor portal
Recording of selected instruments, replay of any fragment of the day, the investor portal and notifications.
-
October 2026
Mobile version and AI tools
A mobile version of the terminal, AI tools for data analysis and for developing the system.
-
ongoing
Regular updates and support
The platform runs live, we keep developing it and are responsible for operating it.
Roles
- We
- Architecture, development (backend, core, frontend, mobile version), operations and support, research on data. No subcontractors.
- Client
- The trader's experience, product decisions and the trading strategy.
[04] How it works
What the system is made of
Integrations
- Broker API: market data and order book streams, orders, operations, portfolio
- Telegram Bot API
- Moscow Exchange trading calendar
Technologies
- Python
- FastAPI
- SQLAlchemy
- PostgreSQL
- Redis Streams
- Celery
- Next.js
- TypeScript
- Custom canvas chart
- Docker
- gRPC
[05] Non-standard
What had to be solved separately
-
Reproducibility
Time is defined by the events of the stream, not by the server clock. So any fragment of the trading day can be repeated and gives the same decisions.
-
Replay through the same code that trades
Recorded data runs through the live code of the core, not a copy of it. A check on a recording shows honestly how the system will behave live.
-
A terminal of our own
The chart, the order book heatmap, the decision journal and rewind through the day are built for a trader's work rather than taken from an off-the-shelf package.
-
Safe switch-on of new logic
New logic is first computed alongside the live one in shadow mode. The interface shows a confirmation from every engine, not the position of a switch.
[06] Screens
What it looks like
[07] Result
What we achieved
The trader sees all instruments and any decision of the system in one terminal and tests ideas on recorded days before going live. Routine work is automated, and investors see their account and receive notifications. We do not publish the trading effect.
- 35 instruments processed at once October 2026 · platform metrics, October 2026
- 5 to 6 million market events per trading day October 2026 · platform metrics, October 2026
- 0.5 ms core processing of a market event, median October 2026 · platform metrics, October 2026
- 23 to 28 ms data delivery from the broker to the core, median October 2026 · platform metrics, October 2026
An honest note on impactThe business impact has not been measured yet, so we quote no percentages or payback times. The figures above describe the project itself.
Delivered
- Everything described in "What we did": the core, recording and replay, the terminal, AI tools, the investor portal, the modes and the mobile version
- Automatic start of the system after a server reboot
Let's discuss a similar task for your business
Discuss a similar task[//] More
More work in this area
A cleaning company platform: orders, specialists and training in one system
The company was entering the Moscow market and wanted to run on its own platform from day one. In a month we launched the MVP: a website with price calculation and ordering, client, specialist and office workspaces, and the training a specialist must complete before taking orders.
Client: Cleanforlife
Taxi fleet management platform
A single system for dispatchers, drivers and managers: orders, contractors, finance and reporting in one place.
Geo-analytics for retail
Analysis of locations and competitive surroundings for choosing points of a retail chain.