[Work] 8 case studies
Systems, explained.
Five case studies are subsystems of one production platform — Astra Raven. Three are tools built around it. Each page shows the architecture, an interactive demo and the decisions behind it.
-
Financial Technology · Backend Engineering · Data Automation
Trading Signal Intelligence Engine
An automated pipeline that reads cryptocurrency market data, narrows a wide universe of instruments down to qualified candidates and turns them into structured, validated trading signals.
Technical focus
- Python
- Async I/O
- Market Data APIs
- Rule Engines
- SQL
Engineering challenges
- Noisy, fast-moving market data
- Multi-stage filtering without losing traceability
- Deterministic output from many inputs
What was built
- Market data ingestion
- Candidate ranking & context detection
- Multi-step signal validation
- Persistence, tracking & reports
Explore case studyArchitecture
- Market Data
- Market Analysis
- Candidate Selection
- Signal Validation
- Risk Checks
- Signal Output
- Monitoring
-
Trading Infrastructure · State Machines · Backend Automation
Risk Management & Trade Lifecycle Automation
A controlled lifecycle for trading operations: every signal is validated against risk rules, tracked through explicit states and closed with a recorded result.
Technical focus
- State Machines
- Python
- Exchange APIs
- SQL
- Risk Validation
Engineering challenges
- Operations that must never skip a state
- External API failures mid-lifecycle
- Risk limits enforced in code, not by convention
What was built
- Risk engine with SL/TP and exposure checks
- Explicit trade states & transitions
- Result checker & reports
- Error and rejection paths
Explore case studyArchitecture
- Incoming Signal
- Validation
- Risk Engine
- Execution / Tracking
- Trade State
- Result Checker
- Reports
-
Product Development · UI Engineering · Backend Integration
Telegram Mini App & Product Interface
The user-facing side of the platform: a Telegram Mini App that brings signals, settings, AI chat, subscriptions and account state into one mobile interface in four languages.
Technical focus
- Telegram Mini Apps
- HTML / CSS / JS
- SVG Charts
- API-driven UI
- i18n
Engineering challenges
- Complex backend state on a small screen
- Telegram WebView constraints on iOS
- Four complete locales
What was built
- Dashboard, signals, AI chat, settings
- Subscription & access logic in UI
- Interactive SVG visualisations
- RU / EN / PL / UK localisation
Explore case studyArchitecture
- Telegram Client
- Auth & Session
- Product API
- UI State
- Interactive Views
-
Artificial Intelligence · LLM Integration · Automation
AI Assistant & Context-Aware Automation
AI components inside the platform that understand user requests in context, while deterministic backend rules keep critical operations validated and safe.
Technical focus
- OpenAI API
- LLM Integration
- Context & Memory
- Structured Output
- Validation
Engineering challenges
- Flexible language, strict operations
- Context from live application state
- Model output that must be validated
What was built
- Assistant integration with product backend
- Context retrieval & memory mechanisms
- Structured output validation
- Monitoring of AI workflows
Explore case studyArchitecture
- User Request
- Context Retrieval
- AI Processing
- Structured Validation
- Backend Action
- User Response
-
Backend Engineering · Databases · Software Architecture
Backend Architecture & Data Persistence
The server architecture that holds the platform together: accounts, signals, settings, states, background jobs, access control, reports and monitoring — evolving without rewrites.
Technical focus
- Python
- Async Processing
- SQL Design
- Background Workers
- Logging
Engineering challenges
- Many subsystems sharing consistent state
- Background work alongside user requests
- Growing the system without rewriting it
What was built
- SQL schema & data relationships
- Async workers & scheduled operations
- Validation & business logic layer
- Event logging & delivery to UIs
Explore case studyArchitecture
- External APIs
- Python Backend
- Validation / Logic
- SQL Database
- Background Workers
- Telegram / Web UI
-
Infrastructure · Server Diagnostics · Monitoring
Astra Server Monitor
Diagnostics and observability tools for Linux servers: a clearer view of system state, services and disk usage, with safe, read-only access to diagnostic functions.
Technical focus
- Linux
- Diagnostics
- Observability
- Read-only Access
- Visualisation
Engineering challenges
- Raw server output is hard to read quickly
- Diagnostics without risky write access
- Trustworthy state after deployments
What was built
- Server state & service views
- Process and service observation
- Read-only Disk Explorer
- Controlled diagnostic access
Explore case studyArchitecture
- Linux Host
- Collectors
- Access Control
- Aggregation
- Dashboard
-
Business Automation · Analytics · Backend Product
TrackLink — Tracking & Marketing Analytics
A system for tracked links and marketing events: campaign links, source identifiers, click and event processing, custom domains and analytics delivered through Telegram tools.
Technical focus
- Python
- aiohttp
- Event Tracking
- SQL
- Telegram Mini App
Engineering challenges
- High-volume redirects that must stay fast
- Attributing events to the right source
- Trustworthy analytics from raw events
What was built
- Tracked links & custom domains
- Click and event processing
- Analytics & reporting interface
- Telegram Mini App integration
Explore case studyArchitecture
- Campaign Link
- Tracking Endpoint
- Event Validation
- Data Storage
- Analytics
- Reporting Interface
-
AI Automation · Business Operations · Product Workflows
Marketing Brain — AI Operations Workflows
AI-assisted workflows for marketing operations: content planning, task decomposition, project context and reporting, delivered through Telegram-based interfaces.
Technical focus
- LLM Workflows
- Task Decomposition
- Context Management
- Telegram
Engineering challenges
- Turning broad goals into concrete tasks
- Keeping project context across sessions
- Useful output, not generic AI text
What was built
- AI-assisted content planning
- Structured task decomposition
- Project context & reusable instructions
- Reports for decision support
Explore case studyArchitecture
- Goal
- Project Context
- AI Decomposition
- Structured Tasks
- Reports
No case studies in this category yet.