Grigorii Galichuk

AI/ML • Senior • 23 years • Samara, Russia • UTC+04:00 • Remote only • Work permits: Russia • English: B1 • Available now
Python Developer
Python
SQL
MLOps & Deployment: 4/10
Active 9 days ago
+7 (927) 7989901
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles

Overview

A pragmatic backend engineer at a senior level who reliably composes async services, schedulers and API clients. The strongest proven skill is designing and implementing resilient asynchronous integrations and orchestration, evidenced by the BaseSweeper abstraction and concrete Aptos/Starknet modules with session caching, retries and honeypot simulation (wallet-sweeper). There is limited public evidence of formal API versioning, migration history, or an operations playbook for deployments and secret management.

Technical skills

Python• Senior
SQL• Junior
Python
Aiohttp
Asyncio
Pydantic
aiogram
SQLAlchemy
Databases
PostgreSQL
Redis
AI/ML
Claude
Cursor
Gemini
Groq
Prompt Engineering
LangChain
OpenAI SDK
LLM• 3y+
Langfuse
DevOps
Docker Compose
Git
Ubuntu
Docker
Rest API
QA
Chrome DevTools• 4y+
Postman
Swagger
Sentry

Timeline

DayLog AI
Backend
Apr 2026 to Present 4 Months

DayLog AI — умная экосистема (Telegram-бот + React Mini App) для ведения и аналитики личного дневника на базе искусственного интеллекта. Пользователь просто рассказывает о дне в свободной форме, а бот структурирует текст, задает коучинговые вопросы и строит дашборды.

Ключевые технические решения:

  • Бэкенд: API на FastAPI, база данных PostgreSQL (асинхронная алхимия SQLAlchemy) и Redis для Rate Limiting.
  • ИИ и Отказоустойчивость: Сложный AI-роутинг с каскадными фоллбэками. Для диалогов используется Mistral (Mistral-Small / Ministral-14B) с переключением на Gemini 3.1 Flash-Lite. Для парсинга метрик и аналитики — Qwen 3.6 27B (через Groq API). Промпты защищены от Prompt Injection.
  • Telegram Mini App: Полноценный фронтенд на React 19 + TypeScript.
  • Безопасность (AppSec): Бесшовная аутентификация через проверку initData от Telegram. Полный отказ от localStorage — JWT-токены хранятся строго in-memory для защиты от XSS-атак.
  • Инфраструктура: Мониторинг ошибок через Sentry, аналитика промптов в Langfuse. Проект полностью упакован в Docker Compose для production-развертывания.
Samara Institute of Railway Transport
Bachelor's Degree • Railway, bridge, and transport tunnel construction
2025 Samara, Samara Oblast
Technical Specialist / Python Developer (Analytics) • Middle
Private Practice • Full-Time
Dec 2023 to Dec 2025 2 Years Samara In office
Provided technical support, automation, and analytics for personal Web3-related projects. Built and maintained Python scripts for parsing websites, working with REST APIs, and automating recurring tasks. Administered infrastructure including proxy/anti-detect profiles and security setup, and performed research/manual QA on external web services with bug reporting.
Python
SQL
Rest API
PostgreSQL
Docker
Redis
Postman
Swagger
LLM
Senior AI/ML Engineer Confidence: High LLM Engineer
A practical LLM-integrated developer at a middle-to-senior level focused on building production Telegram chatbots and CLI LLM tools. The strongest proven skill is engineering async conversational systems with persistence and scheduling - shown by scheduler.py (RedisJobStore, persistent vs ephemeral jobs) and db/queries.py (timezone-aware query logic and savepoint handling). There is little-to-no evidence of model training, experiment tracking, automated evaluation suites, or test coverage in public code.
Model Architecture & Training
2/10
How well models are designed and trained
No custom models or training loops; evidence is prompt engineering and LLM API integration (LangChain/OpenRouter) only.
Evidence
bot/services/prompts.py: SYSTEM_PROMPT and METRICS_SYSTEM_PROMPT (detailed prompt engineering)
AI-Assistant/ai_assistant.py: AIAssistant.__init__ using ChatOpenAI and prompt/chain setup
Data Pipeline & Feature Engineering
3/10
How data is prepared for models
Reasonable application-level data handling: async DB queries, timezone-aware date boundaries, and batched digest generation, but no large-scale ETL or feature engineering pipeline.
Evidence
db/queries.py: check_today_entry and get_last_week_entries (timezone-aware day/week boundaries)
bot/services/scheduler.py: run_global_weekly_digest (batching users, rate-limit pause)
Experimentation & Evaluation
1/10
How results are measured and tested
Almost no experiment or evaluation infrastructure (no W&B/MLflow, no experiment configs); only in-app metric generation prompts exist.
Evidence
bot/services/ai.py: get_ai_metrics signature (metrics retrieval pipeline implied)
bot/services/prompts.py: METRICS_SYSTEM_PROMPT (defines desired metrics output format)
MLOps & Deployment
4/10
How models are shipped to production
Clear operational concerns and deployment patterns: persistent Redis jobstore, async scheduler, safe bot retrieval, DB session patterns and background task management.
Evidence
bot/services/scheduler.py: AsyncIOScheduler with RedisJobStore and get_bot()/set_bot() pattern
main.py: RedisStorage usage and startup/shutdown hooks (init_db, scheduler setup)
Computational Efficiency
3/10
How efficiently computing resources are used
Practical async efficiency and batching (asyncio.gather, job batching, memory vs persistent jobstores), but no low-level GPU/quantization or profiling work.
Evidence
bot/services/scheduler.py: batching in run_global_weekly_digest and use of asyncio.gather with throttling sleep
bot/handlers/diary.py: non-blocking background scheduling and cancellation patterns (schedule_nudge/cancel_nudge)
Research Depth & Innovation
1/10
Depth of research and new ideas
No novel research artifacts, custom layers, or paper implementations; prompts and analytic rules reflect domain understanding but not research-grade innovation.
Evidence
bot/services/prompts.py: DIGEST_SYSTEM_PROMPT and rules (design of analysis prompts rather than algorithmic innovation)
Expertise
Conversational AI & Chatbots• Middle
Industries
Artificial Intelligence• Middle
Technologies
Python• Senior
Redis
Cursor
LangChain
Claude
Groq
SQLAlchemy
Prompt Engineering
OpenAI SDK
Git
Gemini
Ubuntu
LLM• 3y+
Asyncio
Pydantic
Aiohttp
aiogram
Recommendations
  • Develop conversational systems and integrations that require async orchestration, scheduled workflows and LLM prompt engineering (telemetry, jobstores, FSM flows).
  • Extend MLOps/inference tooling around LLM usage - add metrics collection, experiment tracking (W&B/MLflow), and structured evals for prompts/metrics generation.
  • Implement automated tests and CI for handlers, DB queries and scheduler logic; add observability (OpenTelemetry/Prometheus) and error-rate alerting for production robustness.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
daylog-ai-public AI-Assistant
Senior Backend Developer Confidence: High API Engineer
A pragmatic backend engineer at a senior level who reliably composes async services, schedulers and API clients. The strongest proven skill is designing and implementing resilient asynchronous integrations and orchestration, evidenced by the BaseSweeper abstraction and concrete Aptos/Starknet modules with session caching, retries and honeypot simulation (wallet-sweeper). There is limited public evidence of formal API versioning, migration history, or an operations playbook for deployments and secret management.
API Design
3/10
How well APIs are designed
API-level design and client usage are consistent and pragmatic (callback validation, job ids, rate-limit handling), but there is little evidence of formal API versioning, idempotency key patterns, or documented error contracts for public HTTP APIs.
Evidence
daylog-ai-public/bot/handlers/start.py: callback_data validation (process_digest_day/process_digest_time) and reply_markup usage
wallet-sweeper/modules/starknet/sweeper.py: explicit 429 rate-limit handling with Retry-After backoff in get_tokens
daylog-ai-public/bot/services/scheduler.py: cron job registration and job id patterns (schedule_daily_reminder, schedule_global_weekly_digest)
Data Layer & Database
5/10
Working with databases
Thoughtful async DB usage with transaction awareness, flush semantics and race condition mitigation; no migration history or schema evolution files were present.
Evidence
daylog-ai-public/db/queries.py: get_or_create_user uses async with session.begin_nested() and handles IntegrityError for race conditions
daylog-ai-public/db/queries.py: uses session.flush() to push updates without committing, indicating middleware-managed transactions
daylog-ai-public/db/models.py: Declarative models (User, DiaryEntry) with mapped_column and relationships
Scalability & Performance
5/10
Handling load and speed
Clear concurrency and performance considerations: session reuse, connection timeouts, semaphore-based parallelism and batching for external LLM rate limits; caching and pooling shown for HTTP sessions.
Evidence
wallet-sweeper/core/base_sweeper.py: run uses asyncio.Semaphore and _retry implements exponential backoff retry logic
wallet-sweeper/modules/aptos/sweeper.py: _get_session caches aiohttp.ClientSession per proxy and sets ClientTimeout
daylog-ai-public/bot/services/scheduler.py: run_global_weekly_digest batches users with asyncio.gather and sleeps to respect RPM limits
System Architecture
5/10
Overall system structure
Deliberate modular decomposition with abstract base classes, per-network modules, and separated bot handlers/services/middlewares; configuration is centralized, but there is no evidence of multi-service inter-service contracts or a distributed tracing/mesh setup.
Evidence
wallet-sweeper/core/base_sweeper.py: BaseSweeper abstract class with concrete Aptos/Starknet implementations
wallet-sweeper/modules/starknet/sweeper.py and modules/aptos/sweeper.py: clear module separation for blockchain-specific logic
daylog-ai-public/bot: separation into handlers, services (ai, scheduler, saver), middlewares and utils (e.g., bot/handlers/start.py, bot/services/scheduler.py, bot/middlewares/db.py)
Security & Auth
4/10
Protecting data and access
Security-aware patterns are present: masking secrets in reports, honeypot simulation, blacklisting, and cautious signing for simulations; however secrets lifecycle, secure storage, and a dependency security audit are not visible.
Evidence
wallet-sweeper/core/excel_exporter.py: _mask_seed masks seed phrases before exporting
wallet-sweeper/core/base_sweeper.py: blacklist of honeypot contracts and simulate-before-send workflow
wallet-sweeper/modules/aptos/sweeper.py: simulate_transfer uses simulate endpoint and treats signature cautiously for simulation
Reliability & Observability
5/10
Stability and monitoring
Good reliability and observability practices: structured logging, retries with backoff, timeouts, graceful task cancellation and job persistence via Redis; some operational pieces (metrics, alerts, migrations) are not fully shown.
Evidence
wallet-sweeper/core/base_sweeper.py: _retry implements exponential backoff with logging and max retry handling
wallet-sweeper/modules/aptos/sweeper.py and modules/starknet/sweeper.py: aiohttp.ClientTimeout usage and session reuse to avoid connection churn
daylog-ai-public/bot/logging_config.py: custom logging setup (InterceptHandler) and logger binding in services (logger = logger.bind(module="SCHEDULER"))
Expertise
Backend AI & LLM• Middle
Messaging & Real-time• Senior
Python• Senior
Industries
Blockchain & Crypto• Senior
Artificial Intelligence• Middle
Technologies
PostgreSQL
Recommendations
  • Lead development of async API backends that integrate external HTTP services, LLMs and scheduled jobs - implement new connectors or orchestrators using BaseSweeper-like abstractions.
  • Implement or harden production job scheduling and observability (Prometheus metrics, alerts, structured tracing) around the bot and batch digest flows.
  • Design and implement secure secret management and migration workflows - add DB migration history (Alembic) and vault-backed secrets for keys and API tokens.
  • Build real-time messaging features or bot-based products that need FSM, rate-limit handling and durable jobstores (Redis) - continue improving idempotency and error contract practices.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
daylog-ai-public wallet-sweeper AI-Assistant
Senior DevOps Engineer Confidence: Medium Generalist
A generalist senior-level engineer (mid senior, tier_score 4.5) who builds reliable async Python services and blockchain integration tools. The strongest proven skill is designing reliable asynchronous systems and domain-specific blockchain sweep/bridge logic, supported by concrete artifacts such as core/base_sweeper.py and modules/aptos/sweeper.py (key derivation, retries, session caching, simulation logic) and the scheduler with persistent Redis jobstore in bot/services/scheduler.py. Public code shows limited CI/CD, IaC and secrets-management practices - there is little evidence of k8s manifests, secret vaulting, formal SLOs/alerting or production-grade pipeline automation.
CI/CD Pipelines
2/10
Automated build and deploy
Minimal CI/CD evidence - there is a Dockerfile and docker-compose but no pipeline definitions, reusable workflows, artifact handling or deploy gates.
Evidence
daylog-ai-public/Dockerfile
daylog-ai-public/docker-compose.yml
Infrastructure as Code
1/10
Managing servers with code
Very light IaC footprint - only docker-compose appears; no remote state, Terraform/Pulumi modules, environment-separated state, or infra tests.
Evidence
daylog-ai-public/docker-compose.yml
Containerization & Orchestration
2/10
Working with containers
Basic containerization: simple Python Dockerfile and compose deployment. No Kubernetes manifests, health probes, resource rationale beyond a DB memory limit, or pod disruption/config tuning.
Evidence
daylog-ai-public/Dockerfile
daylog-ai-public/docker-compose.yml
Observability & Monitoring
3/10
Watching system health
Observability is lightweight: structured logging and logger wrappers exist, and scheduler logs include batching/rate-limit notes, but there are no SLOs, metrics exports, alert routing or formal dashboards as code.
Evidence
daylog-ai-public/bot/logging_config.py
linea-bridgooor/logger.py
daylog-ai-public/bot/services/scheduler.py
Reliability & Incident Response
5/10
Keeping systems up
Strong reliability and incident-focused engineering for application-level services: retry helpers, concurrency control, session caching/cleanup, nested DB transaction handling for race conditions, persistent jobstore for scheduled jobs and graceful autosave logic.
Evidence
wallet-sweeper/core/base_sweeper.py
daylog-ai-public/db/queries.py
daylog-ai-public/bot/services/scheduler.py
Cloud & Cost Optimization
2/10
Smart use of the cloud
Some cost/rate-aware decisions (batching, pauses for API RPM limits, memory limit on DB) but no systematic cloud cost optimisation, autoscaling strategies, spot/eviction handling or IAM/workload identity patterns.
Evidence
daylog-ai-public/bot/services/scheduler.py
daylog-ai-public/docker-compose.yml
Expertise
Site Reliability Engineering• Senior
Industries
Blockchain & Crypto• Senior
Artificial Intelligence• Middle
Technologies
Docker Compose
Recommendations
  • Use this developer to build and maintain async Python backend services that require scheduling, retry semantics and DB transactional safety (examples: LLM-backed bots, job workers).
  • Assign them to implement blockchain integration tooling where careful on-chain simulation, key-derivation and safe transfer logic are required (examples: sweepers, bridge helpers), while pairing with security reviews for key handling.
  • Have them lead reliability improvements: add observability (metrics, Prometheus exporters, SLOs), structured alerting and incident runbooks based on the existing scheduler and retry patterns.
  • Pair them with a DevOps engineer to add CI/CD pipelines, secrets management (Vault/SOPS), IaC (Terraform/CloudFormation) and k8s manifests for safer production deployments.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
daylog-ai-public wallet-sweeper linea-bridgooor