LMC Tech
Published Android app
LODORi
A meditation app that writes and voices a personalised session in about 25 seconds. React Native, FastAPI, subscriptions, ads, safety checks, and a public release on Google Play.
Our work / Data and finance automation
AI trading research
A research platform for testing whether language-model signals plus strict risk rules can trade well, built to fail closed. The equities bot scans a watchlist every fifteen minutes in market hours through Interactive Brokers' paper gateway; the crypto bot runs around the clock on a testnet and refuses to start if it is pointed at a live exchange.
250+
commits across the two systems
5%
maximum position size, 3% daily loss halt, enforced in code
15min
equities scan cadence in market hours; crypto hourly, 24/7
0
real money: paper accounts and testnet only, by design
The brief
Good trades, not lots of trades.
We wanted to know, with evidence, whether AI-generated signals are worth anything once you apply the risk discipline a real desk would insist on. That means a system that is honest about its results, cannot be talked into a bad position, and keeps running when the broker connection drops at 3am.
- Signal generation with a reflection pass, confidence and market-regime gates, and rule-based fallbacks when the model API is down.
- Hard risk engine: per-position caps, stop and take-profit levels, open-position limits and a daily loss limit that halts trading.
- Broker-truth dashboards with live portfolio, trades, signals and health endpoints; Telegram alerts for the crypto system.
- Self-healing broker layer with an automated re-login watchdog and an incident sentinel that reports to our Slack.
- Backtesting harness with dozens of stored parameter runs and a weekly strategy review job.
- Ops discipline: security audit, incident runbook, secrets model, network exposure register, SSH only over a private network.
Under the bonnet
FastAPI, schedulers and a lot of guard rails.
Python with FastAPI, SQLAlchemy and APScheduler, ib_async against IB Gateway for equities, ccxt for crypto, Gemini with Groq and local Ollama fallbacks for signals, Chart.js dashboards, Docker and systemd on a dedicated VPS with webhook deploys.
- Python
- FastAPI
- SQLAlchemy
- APScheduler
- ib_async
- ccxt
- Gemini
- Ollama
- Docker
- Telegram
What it shows
- Data pipelines, scheduling and dashboards for time-critical systems.
- AI decisions wrapped in deterministic safety rules.
- Resilience engineering: watchdogs, sentinels, runbooks.
- Nothing here is investment advice, and none of it trades real money.
AI social media manager
Social Jen
Give it one idea, a product photo or a video and it drafts, schedules and publishes to nine social networks, writes graded Etsy listings and renders your 3D models. Nothing goes out until you approve it.
AI to 3D print pipeline
modelgen
Describe a part or drop in a picture and get a watertight, print-ready STL. Exact parametric CAD written by a language model, or organic meshes from AI, both validated before a slicer ever sees them.
Next step