РУҚАЗEN中文
async def scan() model.predict(x) ws.connect() fft(signal) agent.run()
Development Portfolio Industrial Automation · AI · Web3 · Cybersecurity

Applied IT systems and engineering complexes

A catalog of 19 applied IT systems developed by Dmitry Schultz and the Neyro Veles team: from mobile SCADA dispatching of a heating boiler plant to multi-agent AI complexes, algorithmic trading, and automated data leak auditing. Every project was built for a specific task — with a working prototype, not just a presentation.

Industrial Automation & IoT

Android SCADA: Mobile Boiler Plant Dispatching

A mobile HMI application for remote supervisory control of a heating boiler plant in a five-story, four-entrance residential building — 60 apartments. The dispatcher monitors the boiler status and controls it from a smartphone, without traveling to the site.

Features
  • Remote start and emergency shutdown of the gas boiler
  • Real-time hydraulic pressure monitoring in the circuit
  • Continuous monitoring of supply and return coolant temperatures
  • Emergency protocol with protective acknowledgment and push/SMS alerts

Result: heating season under 24/7 control — emergencies are visible the moment they occur, not after a resident's phone call; on-call crew dispatches reduced to only truly necessary ones.

SCADA Project Page →
Android · KotlinModbus RTU/TCPOPC UA GatewayWebSocket APIGas Boilers
AI & Multi-Agent Systems

MonAmour Swarm RAG — Multi-Agent AI Consultant

A multi-agent smart perfume and cosmetics consulting system built on the LLM Wiki + RAG principle. Agents are divided by roles: one gathers requirements, another answers from the knowledge base, and a third drives the sale to completion.

Features
  • Role-based agent separation — precise consultations without hallucinations
  • Deep product matching by ingredients, notes, skin type, and budget
  • Answers strictly from the verified knowledge base with source attribution
  • End-to-end operation in Telegram, Instagram Direct, and CRM

Result: expert-level consulting — available 24/7 and simultaneously for all customers, without increasing the payroll.

Python 3.11Swarm / Multi-AgentRAG · Vector DBFastAPITelegram API
AI Infrastructure & LLM Gateway

SmartRouter / OmniRoute — Intelligent AI Gateway

An LLM request router: it decides which model to assign the task to, compresses context before sending, and switches to a backup model if the primary one is unavailable or too expensive.

Features
  • Adaptive context compression (RTK + Caveman) — savings from 15% to 95% of tokens
  • Cascading fallback to free and low-cost models
  • Unified API on top of OpenAI, Anthropic, Bedrock, and open-source models
  • Per-project usage tracking and fault-tolerant AI access

Result: LLM API costs drop dramatically while maintaining response quality, and the service stays up even if one provider fails.

Reduce AI Costs →
Python · FastAPIAWS BedrockOpenAI / AnthropicToken CompressionFallback Cascade
AI Agents & QA Automation

SDET-GENIE — Autonomous UI Testing

AI agents that test web applications just like a human tester: they open a browser, click, fill out forms, verify results — all without manually writing Selenium scripts.

Features
  • Multimodal browser control: the agent "sees" the page
  • Test case generation from plain-language user scenario descriptions
  • Automatic regression runs after each release
  • Report with reproduction steps for each failure

Result: regression testing is no longer a release bottleneck — scenarios are verified in minutes, not a full day of manual work.

Automate Testing →
AgnoBrowser-usePythonVision LLMCI/CD
Web3 & Algorithmic Trading

Polymarket BTC 5-Min — Autonomous Trading Bot

A bot and dashboard for five-minute Up/Down binary options on Polymarket: it compares BTC prices from three independent sources in real time and catches discrepancies between the prediction market quote and the actual price.

Features
  • Simultaneous price feeds from Bybit, Chainlink oracles, and Polymarket
  • Spread detection and delta price change within a round
  • Automatic position entry based on a defined strategy
  • Live dashboard of round statuses and open positions

Result: decisions are made in milliseconds — faster than price discrepancies can be closed manually.

Discuss a Trading Bot →
Python · WebSocketPolymarket APIBybit APIChainlink OraclesRealtime Dashboard
Exchange Analytics & ML

Binance Futures Pressure & Correlation Detector

A Binance futures market analysis system: it calculates Order Book Pressure in real time and discovers hidden correlations between assets using clustering methods.

Features
  • Buyer/seller imbalance calculation by order book depth
  • Asset clustering and discovery of correlated pairs that move together
  • Streaming tick processing with zero delay via WebSocket
  • Signals about shifts in market forces before a price reversal

Result: the market picture is visible before it shows on the chart — from the order book, not from price movement that already happened.

Order an Analytics System →
Python · PandasWebSocketsML ClusteringBinance Futures API
Deep Reinforcement Learning

Crypto RL Modern — Reinforcement Learning Trading

Algorithmic trading based on deep reinforcement learning using limit order books (LOB). The agent does not follow hardcoded rules — it learns the strategy from historical data and is validated on a held-out period.

Features
  • Agent training in an exchange simulator (Gym / Gymnasium)
  • Multi-horizon price movement forecasting
  • Backtesting with honest train/test data splitting
  • Risk and drawdown control as part of the reward function

Result: the strategy is discovered by data, not manually fitted, and its behavior is pre-validated on a period unknown to the agent.

Discuss an ML Task →
PyTorchStable-Baselines3GymnasiumLimit Order BookBacktesting
High-Load Screener & Monitoring

MEXC Volume Intelligence & Crypto Super Screener

A high-load screener that continuously monitors hundreds of trading pairs and instantly highlights anomalous volume spikes — the moment big money enters an asset.

Features
  • Simultaneous real-time monitoring of hundreds of pairs
  • Volume spike detection relative to a moving average baseline
  • Web dashboard with live sorting and instant notifications
  • Flexible trigger thresholds for your own strategy

Result: the signal arrives at the moment of the spike, not after the move has already played out.

Order a Screener →
Python · FastAPIWebSocketsMEXC APIJS Dashboard
Blockchain Analytics

Blockchain Token Analytics — ERC-20 Transaction Monitoring

A server for tracking ERC-20 token transactions on the Ethereum network in real time: it catches large transfers ("whale movements") and feeds them into a dashboard and notifications.

Features
  • Contract event subscription via Ethereum WebSocket
  • Transfer filtering by amount, token, and address
  • Data enrichment via Etherscan API and address history
  • History storage and export for further analysis

Result: large fund movements are visible at the moment of block confirmation — without manual block explorer monitoring.

Discuss Web3 Development →
Web3.pyEthereum WSEtherscan APIERC-20PostgreSQL
Cybersecurity & Secrets Auditing

KeyScan Engine — Automated Data Leak Auditing

A cybersecurity auditing and leaked credential validation system: it finds compromised API keys, access tokens, and database connection strings, checks if they are still active, and assesses criticality.

Features
  • Deep search for keys, tokens, and credentials in leaked databases and repositories
  • Mass asynchronous health-check: which key is still valid
  • Automatic risk criticality assessment and prioritization
  • Secure storage of findings and responsible party notifications

Result: the full "detect → validate → notify" cycle runs automatically, reducing leak response time from weeks to hours.

Order a Security Audit →
Python AsyncioRegex EngineHealth-CheckerRisk Scoring
Sales Automation & Parsing

MegaBrain Upwork & Kwork — Feed Parser & Auto-Response

Industrial parsers for freelance marketplace order feeds with automatic proposal submission: the system monitors new projects 24/7, selects relevant ones, and generates a personalized cover letter via AI.

Features
  • Upwork: works via GraphQL API with bot protection bypass (curl_cffi)
  • Kwork: 24/7 feed monitoring and instant reaction to new orders
  • AI-generated proposals tailored to each specific job description
  • History storage in PostgreSQL and proposal conversion reports

Result: the proposal goes out within the first minutes after job posting — where on freelance platforms, that timing decides who gets hired.

Automate Sales →
Python AsyncioGraphQLcurl_cffiPostgreSQLSmartRouter AI
HR-Tech & Auto-Apply

HH Applicant Tool & AI Search — HeadHunter Automation

An automated job search suite for HH.ru with smart application submission: it filters results by real criteria, and the cover letter is written by AI — tailored to each specific job posting.

Features
  • Automated search and filtering of job listings, including remote positions
  • Cover letter generation tailored to the job requirements
  • Mass application submission with platform rate-limit awareness
  • Vue.js web interface for management and statistics

Result: job coverage increases tenfold with the same time investment, and cover letters are no longer identical copy-pastes.

Discuss HR Automation →
PythonVue.jsPlaywrightHH APILLM
GovTech & Tenders

Goszakup AI — Public Procurement Monitoring & Analysis

A continuous monitoring system for Kazakhstan government procurement with AI relevance scoring: out of thousands of lots, only those the company can realistically bid on remain.

Features
  • Daily collection of new lots from the public procurement portal
  • AI analysis of technical specifications for company profile match
  • Ranking by amount, deadlines, and win probability
  • Notifications about matching lots on the day of publication

Result: no relevant tender goes unnoticed, and the specialist reviews 5 relevant lots instead of scrolling through 500.

Set Up Tender Monitoring →
PythonLLM RelevanceGoszakup APIScheduler
Speech Technologies & NLP

GigaAM Speech Processing — Speech Recognition

Integration of open-source GigaAM acoustic models for fast speech-to-text transcription and multimodal processing — transcribing calls, voice messages, and meetings.

Features
  • High-speed audio-to-text transcription (ASR)
  • Transcription of voice messages from messengers and call recordings
  • Further AI text analysis: key points, tasks, agreements
  • On-premise processing — audio never leaves your servers

Result: conversations are automatically turned into text and structured data — call quality control without manual listening.

Implement Call Transcription →
GigaAMASRPyTorchOn-Premise
Mobile Apps & CRM

Android Apps with Kaspi Pay and CRM Funnels

Commercial Android mobile applications with online payments and end-to-end CRM integration. A classic example — a delivery app for a cafe: orders come in directly, with zero aggregator fees.

Features
  • Your own delivery app instead of an aggregator — 0% commission
  • Payment via Kaspi Pay QR directly in the app
  • 1C-Bitrix24 and amoCRM integration: a lead automatically creates a deal
  • Bitrix24 linked with 1C:Trade Management — invoices are issued automatically
  • Push notifications and loyalty programs for repeat orders

Result: the aggregator's commission stays in your business, and the customer base belongs to the owner, not the platform. Apps from 450,000 ₸, Bitrix24 implementation from 350,000 ₸.

Learn More About Development →
AndroidKaspi Pay API1C-Bitrix241C:TMamoCRM
Quantitative Analytics & ML

Quantitative Signal Analysis — Crypto Market Signal Decomposition

A research engine for quantitative analysis of price time series: Fourier decomposition, wavelets (PyWavelets), EMD, PCA feature decomposition. The goal is to uncover hidden cycles and patterns invisible on a bare chart.

Features
  • Spectral analysis (FFT, Welch) — extracting dominant price movement frequencies
  • Wavelet decomposition and Empirical Mode Decomposition (EMD)
  • PCA + multicollinearity: automatic feature noise cleaning
  • Asynchronous batch processing of thousands of addresses and metrics

Result: stable cyclic components are extracted from a chaotic chart — input data for trading models, not intuition.

Order Analytics →
Python · NumPySciPy FFTPyWaveletsPyEMDScikit-learn PCA
Multi-Exchange Trading & Pump Detection

Pump Detector & Multi-Exchange Bandits (MEXC, Binance, Bitget)

An ecosystem of trading bots operating simultaneously on MEXC, Binance, and Bitget: instant pump movement detection and algorithmic position entry ahead of the crowd. A "bandit" approach — the bot adapts its strategy to each exchange in real time.

Features
  • Pump signal detection by volume anomaly and price change velocity
  • Parallel operation on 3 exchanges with a unified control panel
  • Adaptive "bandit" algorithm: the bot reinforces the winning strategy
  • Drawdown control and automatic stop when limits are exceeded

Result: pumps are caught in seconds — while most participants are just learning about the movement, the bot is already in position.

Discuss Exchange Bots →
Python · CCXTMEXC APIBinance APIBitget APIMulti-Armed Bandit
Web3 Automation & Airdrop

HotWalletBot & ClaimClicks — Airdrop Automation

A suite for mass claiming of crypto airdrops and faucets: multi-threaded operation with dozens of wallets simultaneously, proxy rotation, anti-bot bypass, and automatic thread distribution.

Features
  • Parallel claiming via N threads with a dedicated proxy for each
  • Hot wallet pool management and automatic fund collection
  • Cloudflare and CAPTCHA bypass via nodriver / undetected-chromedriver
  • User-Agent, fingerprint, and proxy server rotation

Result: hundreds of claims per hour instead of manual processing — a task that would take a day is handled automatically.

Automate Web3 →
Python · nodriverMultiprocessingProxy RotationAnti-Bot BypassHot Wallet
Telegram Bots & Knowledge Bases

DaNang AI Guide — Bot with Self-Updating Knowledge Base

An intelligent Telegram bot city guide that automatically collects its knowledge base from topic-specific channels, filters out duplicate posts, and answers users based on accumulated data.

Features
  • Auto-collection and NLP processing of content from Telegram channels
  • Deduplication: the same post never enters the database twice
  • Answers from the accumulated knowledge base with source links
  • The knowledge base grows without administrator involvement

Result: a reference system that never becomes outdated — content updates itself, with no manual entry needed.

AI Assistant for Business →
python-telegram-botSQLAlchemyNLPDeduplication

Core technology stack

Python 3.11FastAPIAsyncioPyTorch Reinforcement LearningRAG · Vector DBOpenAI / AnthropicLangChain · Swarm WebSocketsGraphQLPostgreSQLVue.js Kotlin · AndroidModbus · OPC UAWeb3.pyDocker 1C-Bitrix24amoCRMKaspi PayPlaywright

Need a custom system built for your task?

We provide a free 30-minute express audit, select the tech stack, and deliver a fixed estimate within 24 hours. We work under contract with LLCs and sole proprietors, with electronic invoicing.

Message on WhatsApp