đź“š Tom's Notes

Search

SearchSearch

Trademill, an investment assistant

Jan 22, 20255 min read

An investment assistant in your pocket and in the cloud:

  • Always monitoring your existing investments for risks and surprises
  • Periodically scanning the markets for new opportunities
  • Suggesting new investments, based on your criteria, when capital is available
  • And executing trades upon user approval

Business model

  • Subscription-based

  • Features for premium subscription?

    • redundant / HA setup
    • voting logic before executing trades

Product

Jobs to be done:

Understand and evaluate this product

Landing page with images

Demo env

  • pull pre-baked data or rely on paper trading account ?
  • set up strategy
  • replay series
  • show events

Set up account and fund

Identify investments

  • Start from the universe of all investments
  • Narrow down by sector, theme, performance
  • Compare alternatives/candidates/companies
  • Invest
  • Let assistant monitor your investments and notify of significant changes

Know when conditions “change"

"Change” has different meaning for each investor, need to be customizable

Features

Whole-market metrics

  • Change in total capitalization
  • Shiller P/E
  • Total earnings
  • Total P/E
  • Total FCF ??

Historical prices

  • zoomable OHLC chart

Example: https://github.com/leeoniya/uPlot/blob/master/demos/candlestick-ohlc.html

Investment strategies

  1. Store rules locally
  2. Edit at the bottom of chart (for now)

Positions

New opportunities

  • Screener (external system? ) sending updates to this component
  • Opportunities listed in main page with overall score
  • Can be expanded with quick company summary and key metrics

Fundamentals view

  • usual stats (PE, PS, …)
  • x-y chart with cash flow and growth
    • background of chart shows dots for total market
    • hovering on dots show some representative companies (find a proxy for “well known”)

Editing rules

  • each rule had a unique ID. UUID v4 for simplicity
  • upsert_rule() method to submit the single new / modified rule to the engine
  • check_rule() method to test validity and report to UI any issues

UX awesomeness

  • For every rule, calculate how close it is to trigger. Show indication in UI

User / group notifications

  • Send notification when a rule triggers
  • and when a trade executes
  • Summary message at end of day?

https://ajlopez.wordpress.com/2015/02/06/enviando-mensaje-a-slack-con-javascriptnode-js/

Web UI to monitor and edit strategies

Mobile app for simplified view and critical notifications

Strategies to support

Globally-optimized portfolio

See separate note

Broad market, dynamic allocation

  • Change the allocation between equities and cash equivalent based on Schiller P/E value

Asset class rotation

https://novelinvestor.com/asset-class-returns/

Sector rotation

https://novelinvestor.com/sector-performance/

Buy quality + value

Whenever new funds available in the account, above a threshold:

  • Filter companies by score
  • Invest in the N most undervalued (by P/E, below a threshold)

Hand-pick investments

  • Search companies
  • Filter by fundamental and technical indicators
  • Batch buy

Legal

Disclaimer : not to be construed as investment advice, investment strategies are chosen by the user

Liability

Determine liability due to:

  • Incorrect market data
  • Failure to notify user
  • Failure to execute trade after it’s been approved by the user

Architecture

JITA : Just-In-Time Architecture

  • Enough complexity to handle the current product’s use cases.

  • Plan for evolution but don’t over-engineer

  • Start with web UI, not mobile

  • Data pipeline and analytics DB are co-located until that causes performance bottleneck and/or availability risk

Longer term

  • Run on isolated Firecracker VMs, one per user (single tenant ⇒ higher security)
  • Support subscription model

Open issues

  • management of secrets / keys ?
    • solution from cloud provider. See how it can be tied to auth / identity service

Tech - Web UI

svelte UI - webview

https://developer.chrome.com/docs/multidevice/webview/gettingstarted/

eCharts

  • time series
  • charts with confidence bands
  • radar (esp. when comparing)
  • sankey
  • streak on calendar view
  • tree map

https://echarts.apache.org/v4/examples/en/index.html

Svelte, data table and store

https://italo-silva.medium.com/svelte-how-to-create-a-dynamic-table-component-1c7d33a307e3

Tech - App backend

  • Per-user
  • Running in a microVM (Firecracker-based)
  • With user profile DB (local Sqlite) replicated off-site
  • With company-data DB shared across instances

Scalable, fully-isolated infra

https://fly.io

with scale to zero and fast starts

https://fly.io/blog/fly-machines/

Sortable unique IDs

  • UUID v7

https://buildkite.com/blog/goodbye-integers-hello-uuids

https://github.com/segmentio/ksuid https://github.com/novemberborn/ksuid

Mostly-read-only DB for stock screener

https://www.sqlitetutorial.net/sqlite-nodejs/query/

Tech - data pipeline

Start simple, make it progressively more complex

  • Metaflow running locally

  • Metaflow on single instance, where pipeline is triggered by GitHub Actions

  • GH Actions also used to drive acceptance test (same pipeline processing a subset of entities and writing output to <branch> instead of prod / main destination)

  • Ingest into analytics DB only what has changed from previous run

Tech - future mobile UI

Flutter?

Candlestick (syncfusion) chart

https://www.syncfusion.com/flutter-widgets/flutter-charts/chart-types/candle-chart

Local and push notifications

https://pub.dev/packages/awesome_notifications

Local notifications

https://blog.logrocket.com/implementing-local-notifications-in-flutter/

Home widget

https://pub.dev/packages/home_widget/example

Svelte + Capacitor?

https://github.com/drannex42/svelte-capacitor

  • YUCK: uses whole Android Studio to build pkg
  • 16GB machine freezes during build

Android only, Kotlin?

App + websockets + chart

https://github.com/MohitMandalia/AirQualityVisualizer

https://github.com/therajanmaurya/Stock-Hawk

Mobile home screen widget

https://github.com/premnirmal/StockTicker

native UI - charts

https://github.com/AnyChart/AnyChart-Android

https://github.com/AnyChart/AnyChart-Android/blob/master/sample/src/main/java/com/anychart/sample/charts/OHLCChartActivity.java

notifications with locked screen

https://developer.android.com/guide/topics/ui/notifiers/notifications

Data sources

http://www.columbia.edu/~tmd2142/best-6-stock-market-apis-for-2020.html

Robinhood realtime data

https://github.com/sanko/Robinhood

Tradier

https://documentation.tradier.com/brokerage-api

Rate limits: https://documentation.tradier.com/brokerage-api/overview/rate-limiting

TradingView

https://github.com/Mathieu2301/Tradingview-API

Misc APIs

https://rapidapi.com/collection/tradingview-api

FinViz

Finnhub.io

Simfin

Https://SimFin.com Company fundamentals

IEXCloud

Deployment

Serving DL models

https://github.com/pytorch/serve

https://aws.amazon.com/blogs/machine-learning/deploying-pytorch-models-for-inference-at-scale-using-torchserve/

https://medium.datadriveninvestor.com/deploy-your-pytorch-model-to-production-f69460192217

Launch

Blog posts to introduce project

What is it?

  • Investment assistant
  • Solo developer
  • Calm SaaS
  • What we’ll cover next (link to 2 articles below as they are published)

Data pipeline

  • Design goals: low cost, high developement velocity
  • Overview
    • Ingestion
    • Analytics DB
    • Support for data science

App and trading engine

  • Design goals: robust, high dev efficiency
  • Overview
    • Frontend, Svelte
    • Backend, data streams, realtime updates, trade execution

Resources

Graph View

Backlinks