Fragmented workflow
Market context, risk calculation and execution planning normally live across separate tools and screens.
AI-assisted product case study · 2026
UCB°
TRADING
A Telegram bot and Mini App that turns live MEXC data into structured trading plans with personalised risk controls.
I designed and built the complete product—from onboarding and market analysis to subscriptions, payment verification and a responsive Mini App.
40Passing automated tests
20/20Live markets processed
100%Generated plans valid
3.18sMedian full analysis
01 / Problem
FROM MARKET NOISE
TO A CLEAR PLAN
Risk is personal
The same signal means different exposure for every user. Deposit, risk tolerance, leverage and margin mode must shape the output.
Speed without opacity
The product needs to deliver a fast answer while keeping entry, stop, targets and required margin visible and understandable.
02 / User flow
ONE PATH,
FIVE STEPS
Start in Telegram
Choose a language and enter the product without a separate account.
Set risk profile
Save deposit, risk percentage, leverage and margin mode.
Analyse market
Fetch MEXC data and combine structure, momentum and volatility across timeframes.
Build a plan
Calculate entry, stop, TP1/TP2, position size and required margin.
Review and track
Open the Mini App for charts, settings and persistent signal history.
03 / Product
THE ANALYSIS
WORKSPACE
04 / System
MORE THAN
A PROMPT
The analysis engine is deterministic. AI-assisted development helped me move from product hypothesis to working software, but the production output comes from explicit market logic, validation and risk formulas.
Telegram useronboarding · commands · alerts
↓
aiogram bot + FastAPI Mini Appinterface · API · access
↓
MEXC + analysis enginelive data · MTF structure · confidence
↓
Risk and position sizingentry · stop · targets · margin
↓
PostgreSQLprofiles · signals · subscriptions · deduplication
05 / Decisions
PRODUCT CHOICES,
NOT JUST FEATURES
Remove account friction
Telegram provides identity, distribution and notifications in one place. The Mini App adds visual depth without forcing users into a separate product.
One signal, personal exposure
Market analysis is stored independently from user-specific position sizing, so the same setup can be translated into an individual risk plan.
Make outputs inspectable
Structure, momentum, volatility and confidence are calculated through explicit rules rather than an opaque generative answer.
Design monetisation into the flow
Trial limits, subscription state and confirmed USDT TRC20 payments are part of the same product journey—not a later mock-up.
06 / Build timeline
SHIP FAST.
THEN NARROW.
Signal-bot MVP
A deployable Telegram bot, MEXC data, scanners, multi-user commands and a five-timeframe analysis engine.
Autotrading expansion
BingX execution, grids, leverage controls and multiple strategies exposed growing operational complexity.
Strategy narrowing
Liquidity-sweep logic, stricter guards and fewer assets shifted the focus from more trades to safer filtering.
Back to user value
The production MEXC signal experience returned, followed by better explanations, alerts and trial access.
Decision-support product
The codebase consolidated around deterministic plans, personalised risk, five languages, Mini App, payments and persistent state.
Verification pass
README, CI, 40 tests, a 20-market evaluation, product walkthrough and public case study turned the build into inspectable evidence.
Postmortem
AI-assisted speed created breadth. Testing, deletion and scope correction created the product.
07 / Reliability
BUILD,
CHECK, SHIP
- 40 automated tests on Python 3.9
- 20/20 predefined markets processed with fresh MEXC data
- 10/10 generated plans passed every mathematical invariant
- 36/36 personalised risk scenarios passed
- Telegram Mini App HMAC signature validation
- Market symbol and timeframe input validation
- Confirmed on-chain payment and wallet checks
- Persistent alert deduplication in PostgreSQL
- GitHub Actions for compilation and test runs
08 / Live product
START WHERE
USERS ALREADY ARE
The production onboarding supports English, Russian, German, French and Spanish, then routes users directly to deposit settings or the dashboard.