Projects that set new standards for the industry

Task. The client wanted a comprehensive solution uniting all aspects of crypto operations: storing, exchanging, and spending funds through familiar payment instruments. The main challenge was creating a seamless UX so that crypto assets could be spent like regular money while meeting all regulatory requirements.

Technical solution. We designed and delivered a multi-component ecosystem:

Crypto-fiat wallet with cards:

• React Native mobile apps for iOS and Android with custodial asset storage
• EMI integration to issue physical and virtual MasterCard cards
• Apple Pay and Google Pay integration
• Priority debit rules with configurable logic (e.g., USDT → ETH → BTC)
• Conversion module with real-time rates and minimal spreads
• Accounting, reporting, and compliance per European VASP standards

Crypto acquiring for merchants:

• Platform that auto-swaps accepted crypto into stablecoins at transaction time
• Fine-tuned rules and whitelists for coins and networks
• Merchant payout system
• SDK and plugins for popular CMS (WordPress, Joomla, etc.)
• React Native POS terminals for offline retail
• KYT (Know Your Transaction) integration for risk checks

Spot exchange as a liquidity core:

• Centralized liquidity pool serving both wallet and acquiring
• Java matching engine handling thousands of orders per second
• APIs for internal services and external integrators

Architecture highlights:

• Microservices with asynchronous messaging via queues
• Fault-tolerant design with replication of critical data
• PCI DSS standards compliance for card processing

Tech stack: Java, C#, React, React Native, Postgres.

Measurable results:

• Monthly wallet turnover exceeded €1,000,000
• Cards worked worldwide wherever MasterCard is accepted
• Real user transactions processed without critical incidents

Task. Build an institutional-grade crypto exchange for the ADGM (Abu Dhabi Global Market) capable of up to 300,000 transactions per second. Key requirements included full regulatory compliance, listing of any asset class, and direct integration with UAE banking.

Technical solution.

High-performance trading engine:

• Go-based architecture optimized for sub-millisecond latency
• Event-driven order processing achieving 300k TPS throughput
• Support for spot and derivatives
• Ready for horizontal scaling

Regulatory perimeter:

• Full documentation prepared for MTF (Multilateral Trading Facility) and Custodian licenses
• AML/KYC with integrations to international sanctions lists
• Real-time reporting module for the regulator
• Cryptographically protected audit trail

Banking integrations:

• Direct API connections to UAE banks (including SEBA Bank) for fiat operations
• Automatic reconciliation of balances and transactions
• Instant funding/withdrawal system

User interfaces:

• React web platform with advanced trading tools
• .NET MAUI mobile apps for cross-platform delivery

Tech stack: Go, C#, React, .NET MAUI

Results:

• Platform fully engineered and load-tested up to 300k TPS
• Regulatory documentation prepared to obtain early UAE MTF and Custodian licenses
• Architecture ready for horizontal scale under any load
• Banking API integration tests successfully passed

Task. dApp developers and analytics teams needed fast access to structured data across multiple chains. Existing solutions (Moralis, QuickNode) were expensive and lacked the flexibility needed for data processing. We set out to build a customizable, high-performance alternative.

Technical solution.

High-throughput indexing:

• Parallel scanning of multiple EVM chains (Ethereum, BSC, Polygon, Arbitrum) and Solana
• Automatic decoding of smart-contract events using ABIs
• Intelligent retry system on RPC failures
• Reorg detection with automatic data correction

Scalable storage:

• ClickHouse as the primary store for fast analytics
• Partitioning by time and network
• Automatic compression and archiving of older data
• Replication of critical datasets

High-performance API:

• REST API with complex filters and aggregations
• JSON-RPC for compatibility with existing tools
• WebSocket for real-time transaction and event alerts

Analytics capabilities:

• Pre-computed metrics (TVL, trading volumes, top holders)
• OHLCV and trades for all listed tokens
• DeFi protocol tracking with decoding of protocol-specific events

Tech stack: Python (orchestration), Go (critical components), Rust (low-level processing), ClickHouse (storage), Docker + Kubernetes (deploy).

Results:

• Indexing of major chains processing millions of events daily
• 99.5% uptime over the last 6 months

Task. In 2018–2019, Ethereum couldn’t deliver the throughput required for heavy dApps (~15 TPS limit). We needed to build our own L1 with high throughput, fast finality, and a native DEX to provide liquidity.

Technical solution.

Consensus & architecture:

• Delegated Proof-of-Stake (DPoS)
• Throughput: 10,000+ transactions per second
• Fast finality and high fault tolerance

Built-in DEX:

• Native exchange module at the protocol level
• Built-in support for swaps and trading
• Trading fees embedded into the chain’s economy

Wrapped-asset system:

• Vault contracts to mint wrapped versions of external assets (eonBTC, etc.)
• Wrapped-token functionality “out of the box”
• Integration with external chains for bridge creation

Developer tooling:

• SDKs for JavaScript, Python, and Go
• Block explorer with detailed analytics
• Testnet for dApp development and testing
• Documentation and integration examples

Tech stack: C++

Results:

• Blockchain created and successfully launched
• Network operates stably at the stated performance
• An ecosystem formed around EON, including Exscudo/Nimera

Task. Launch an MVP of a p2p crypto exchange under tight deadlines and budget, functionally similar to Garantex.

Technical solution.

Hybrid trading model:

• Platform combining p2p and spot trading
• Custom limit orders linked to external quotes
• Automatic tracking of Binance price with a configurable markup (e.g., Binance +2%)
• Simple architecture with DB-backed matching

Business logic:

• Rules for exchangers and market makers
• Friendly UI for pros who know their markup
• Support for multiple cryptocurrencies and fiat operations

Tech stack: Node.js, React.

Results:

• Production launch within 2 months of receiving the spec
• Platform operated successfully for over a year

Task. Build a closed-loop crypto-fiat ecosystem uniting an exchange, wallet, and acquiring with maximum integration between components.

Technical solution.

Integrated ecosystem:

• High-performance crypto exchange
• Mobile wallet-messenger with encryption via the Signal protocol
• Crypto-fiat acquiring with a shared liquidity pool
• Wrapped fiat tokens backed by funds in bank accounts

Unique traits:

• Advisor team included Moxie Marlinspike, creator of Signal
• Messenger built on the proven Open Whisper protocol
• A unified ecosystem bridging trading, payments, and communication

Tech stack: Java, Swift (iOS), Kotlin (Android).

Results:

• ~2500 BTC raised at ICO in May 2017
• Technically functional, tightly integrated ecosystem

Task. Build a non-custodial collateral protocol for EVM-compatible networks with mechanics similar to AAVE and Compound, with a path to multi-chain deployment.

Technical solution.

• Smart-contract suite enabling collateralized lending
• Web interface for protocol interaction
• Indexer integration for metrics and protocol health monitoring
• Architecture ready for deployment to Ethereum, BSC, Polygon, and other EVM chains

Tech stack: Solidity, React.

Results:
MVP fully tested and ready for deployment

Task. A large international oil & gas company faced heavy loads on HR and accounting due to repetitive requests: certificates, vacations, sick leave, reimbursements, internal procedures. Employees waited for answers; specialists spent time on routine instead of strategic work.

Technical solution.

NLP classification & intent understanding:

• AI chatbot answering organizational questions
• Assistance with document requests and course/training bookings
• NLP-based classification of question types
• Continuous self-learning from user feedback

Enterprise integration:

• Connected to existing HR and accounting systems
• Multi-channel access across corporate platforms
• Knowledge base built from internal policies and procedures

Tech stack: Rasa, DeepPavlov.

Measurable results:

• Monthly HR requests reduced by 50%
• Accounting queries down by 15%
• Specialists focused on more complex, strategic tasks
• Employees got instant answers to common questions

Task. Build a platform for text-based roleplay with AI as the universal game master. Key challenges: thousands of concurrent users, long-session context retention, character/location image generation, and balancing response quality with speed.

Technical solution.

Architecture for high load:

Massive Scale variant:

• Apache Kafka as the central message broker for 10,000+ concurrent users
• Qdrant vector DB for session context and relevant retrieval
• Dedicated GPU cluster for LLM requests

Medium Scale variant:

• RabbitMQ for messaging (up to ~1,000 concurrent users)
• PostgreSQL with pgvector for vector search
• Simple deployment on one or several servers

Model orchestration:

• Automatic model selection based on dialogue context
• Routing between lightweight (simple replies) and heavy models (complex scenes)
• Support for uncensored models for specific requests
• Caching layer to optimize performance

RAG for consistency:

• Vector knowledge base of characters, locations, and world rules
• Semantic retrieval for each AI reply
• Memory management for long sessions

Image generation:

• Stable Diffusion pipeline with LoRA and ControlNet for quality and control
• Upscaling and post-processing pipeline
• Caching of generated images for reuse

Telegram integration:

• Rate limiting to prevent spam
• Inline keyboards for actions and choices
• Payments via Telegram Stars for premium features

Tech stack: Python (FastAPI, aiogram), Kafka/RabbitMQ, Qdrant/PostgreSQL+pgvector, Docker + Kubernetes, Redis, Stable Diffusion, various LLM APIs and local LLMs.

Results:

• Two client profiles implemented for different load expectations
• Massive-scale variant handles thousands of concurrent users
• Demonstrates how to build scalable AI systems

Task. Analysts and portfolio managers spend hours combing through public filings (10-K, 10-Q) hunting for key information across hundreds of pages. The goal: a tool that answers concrete questions instantly with precise citations.

Technical solution.

Automated document processing:

• Automatic download from the SEC EDGAR database
• Conversion to Markdown while preserving structure
• Logical chunking into sections and fragments
• Vectorization for fast retrieval

RAG tuned for finance:

• ChromaDB vector store optimized for financial terms
• Hybrid retrieval for maximum relevance
• Citation system with exact links to source documents
• Company screening by user-defined parameters

Pro analyst interface:

• Chat interface for “talking” to company filings
• Support for complex analytical queries
• Exportable results with source links preserved

Tech stack: Python, ChromaDB.

Results:

• Lightweight service for analysts and investors (a simplified fiscal.ai-style tool)
• Dramatic reduction in time spent on filings
• Instant answers with precise source citations

Task. Automate and speed up marketing content production for a furniture/retail business without expanding the copywriting team.

Technical solution.

• Knowledge-base upload (articles, docs, catalogs)
• Catalog import of product images (furniture, interiors)
• Automatic content generation by topic and tone:
• Titles and copy for posts
• Relevant image selection from the catalog
• Short video and animation generation
• Image-generation pipeline inserting client products into chosen interiors
• Preview and export of ready-to-use marketing visuals

Tech stack: Python, Streamlit.

Results:

• Deployed into the content team’s daily workflow
• Prep time for marketing materials reduced drastically
• Sustained high publishing cadence without new hires

Task. Automate the search and assembly of complete technical documentation for tender participation in the specialized geodesy domain, accounting for the nuances of different work types.

Technical solution.

Intelligent document processing:

• OCR of images and PDFs to text
• AI auto-tags and classifies uploaded documents
• “Foldering” system that sorts by document type
• Tracking of document expiry dates
• Automatic reminders to refresh documents

RAG for corporate search:

• Retrieval across the entire corporate repository
• Works with typed queries and pasted text from other documents
• Semantic search to find related information
• Assembly of complete document packs for specific tenders

Tech stack: Python, Streamlit.

Results:

• Order-of-magnitude reduction in time to prepare tender packs
• Errors from incomplete documentation nearly eliminated
• Transparent status control for each pack

Task. Dairy farms struggled to track many interrelated processes: feed composition and consumption, individual animal metrics, insemination planning, milk yields, veterinary procedures. Excel and paper logs caused accounting errors and profit loss. Western alternatives were unavailable for deployment.

Technical solution.

Implemented modules:

• Feeding analytics: ration planning and nutrition control to standards
• Feed warehouse accounting: stock control with auto-reorder alerts
• Herd movement ledger: full lifecycle from birth to disposal (birth, growth, insemination, lactation, disposal)
• Fuel/equipment accounting with preventive maintenance planning
• Veterinary records with medicine and procedure control
• Integration with state reporting systems

Analytics capabilities:

• Tracking of key production indicators
• Analysis of feeding efficiency and its impact on yields
• Breeding planning based on genealogical data

Tech stack: Python.

Results:

• Successfully deployed on farms totaling >3,000 head
• Significant improvement in process control
• Increased efficiency in planning and resource management

Task. On a crowded real-estate market, the client needed a standout service. The core idea: buyers should see not only the property, but every contractor involved in building it — from architects to finishers.

Technical solution.

Unique cataloging system:

• Structured database of properties with technical specs
• Profiles of all project participants: architects, designers, builders, suppliers
• Ability to find contact details for all contractors on a given property
• Company portfolios with project examples

Advanced search features:

• Interactive map search with filters
• Multi-level filtering by various parameters
• Search by specific contractors or architects
• Ratings and reviews

Tech stack: Ruby on Rails, Angular

Results:

• Launched in 4 months of development
• Client sales up 7% after launch
• Deal-closing time down 39%
• Unique positioning on the real-estate market

Task. Automate and simplify preparation of the complex project documentation farmers need to obtain grants and loans from AO Rosselkhozbank.

Technical solution.

“AgroCifra” financial status report:

• Online project-document builder per AO Rosselkhozbank standards
• Step-by-step data entry with validation
• Automatic generation of all forms and calculations

Tech stack: PHP, Python, Vue

Service for farmer grant programs:

• Document builder for state-support applications
• Full compliance with bank and government standards
• System checks for completeness and correctness
• Tech stack: Node.js, React

Results:

• Significant acceleration of grant-application prep
• Fewer document returns for rework
• Higher approval rates for farmer financing
• Standardized, bank-aligned document workflows