Maksim Zyrianov
Python Developer
9+ years exp
7+ years ML exp
20+ projects
Dart
Objective-C
C
Node JS
SQL
JavaScript
C++
Python
Java
Kotlin
Swift
TypeScript
UX & Visual Polish: 6/10
UI Component Architecture: 5/10
Responsive & Cross-browser: 5/10
Active 7 days ago
+7 (914) 9761825 Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A versatile fullstack developer with solid production experience who produces well-structured, pragmatic code at a middle-to-senior level. The strongest proven skill is building data-driven interfaces and tooling that bridge backend models and live UIs - for example the Polymarket dashboard and the weather forecasting module (forecastDistribution, bandProb, caching and TTL logic). Public code lacks automated a11y audits, measured performance reports, and broad cross-team architecture evidence (multi-service system design), so those areas are not proven here.
Technical skills
Dart
Objective-C
C
Node JS• Middle
SQL• Senior • 9y+ • 10+ projects
JavaScript• Senior • 9y+ • 10+ projects
C++• Senior • 6y+
Python• Senior • 6y+
Java• 6y+
Kotlin• 6y+
Swift• 6y+
TypeScript• Middle • 3y+
Node JS
Dotenv
C++
PyTorch C++
STL
Python
Alembic
Flask
Pydantic
SQLAlchemy
Requests
FastAPI• 7y+
Asyncio• 5y+
Celery• 5y+
Django• 5y+
Databases
Apache Kafka
MySQL
PostgreSQL• 5y+
RabbitMQ• 5y+
Redis• 5y+
AI/ML
LangChain
RAG
Jupyter Notebook
LLM• 7y+
PyTorch• 7y+
Fine-tuning• 4y+ • 10+ projects
Time Series Forecasting• 5+ projects
AI/ML• 5+ projects
Claude• 5+ projects
Gemini• 4 projects
Frontend
Next.js
Tailwind CSS
Vite
Zustand
Zod
React.js• 5y+ • 20+ projects
DevOps
WebSockets
AWS• 9y+
gRPC• 7y+
Rest API• 6y+
CI/CD• 5y+
Docker• 5y+
Mobile
Clean Architecture
React Native• 9y+
Flutter• 6y+
Cybersecurity
IDA Pro• 10y+ • 20+ projects
Timeline
Senior Reverse Engineer
•
Senior
Freelance
•
Freelance
Performed reverse engineering of closed-source Windows binaries and analyzed DLL/EXE files without source code. Reconstructed C++ application architecture, data structures, and class logic via disassembly and decompilation. Developed Python-based tooling to automate analysis and implemented compatible replacements based on observed binary behavior and component interaction.
C++
IDA Pro
Python
Senior Python Backend Developer
•
Senior
Freelance
•
Freelance
Built backend solutions for international SaaS, AI, and business process automation. Developed REST APIs with FastAPI and Django, designed microservice architectures, and implemented async services using AsyncIO. Designed PostgreSQL schemas, integrated Redis for caching/sessions, and built background jobs with Celery and RabbitMQ. Implemented JWT and OAuth2 authentication, integrated external APIs/payment systems, and set up Docker-based deployments with CI/CD.
FastAPI
Django
Asyncio
PostgreSQL
Redis
Celery
RabbitMQ
Docker
CI/CD
Mobile Application Developer
•
Middle
Pavago LLC
•
Full-Time
Developed high-performance native and cross-platform mobile applications for Android and iOS. Worked with Java, Kotlin, Swift, React Native, and Flutter to implement scalable features and integrate REST APIs and third-party services including authentication, payment gateways, and push notifications. Collaborated in cross-functional teams to maintain code performance, security practices, and long-term maintainability.
Java
Kotlin
Swift
React Native
Flutter
AWS
Rest API
AI/LLM Specialist
•
Middle
DrapeFit Inc
•
Full-Time
Built a multi-layer AI recommendation system for personalized style matching at platform scale. Implemented a scoring module using PyTorch with distributed training across multiple GPUs. Improved an AI chatbot by combining Rasa workflows, LLM fine-tuning, and backend integrations using FastAPI and gRPC.
PyTorch
FastAPIsince 2019
gRPC
LLM
Software Engineer
•
Middle
Value Applications
•
Full-Time
Developed a kindergarten management system (Kindy.jp) using React Native and a MERN-based approach deployed on AWS. Supported product rollout to kindergartens in Tokyo and adaptations for the Japanese market requirements. Converted business requirements into scalable technical architecture and collaborated on integration efforts.
React Nativesince 2017
AWSsince 2017
Web Developer
•
Middle
Self-employed
•
Freelance
Implemented custom web applications for Japanese clients through CrowdWorks.jp with focus on responsive UI and performance optimization. Coordinated remotely with clients to translate business requirements into scalable technical solutions. Delivered freelance projects with fixed pricing and hourly models while meeting strict deadlines.
Far Eastern Federal University
Bachelor's Degree •
Information Technology
Middle Frontend Developer
Confidence: Medium Fullstack
A versatile fullstack developer with solid production experience who produces well-structured, pragmatic code at a middle-to-senior level. The strongest proven skill is building data-driven interfaces and tooling that bridge backend models and live UIs - for example the Polymarket dashboard and the weather forecasting module (forecastDistribution, bandProb, caching and TTL logic). Public code lacks automated a11y audits, measured performance reports, and broad cross-team architecture evidence (multi-service system design), so those areas are not proven here.
UI Component Architecture
5/10
How interface parts are built
UI Component Architecture shows custom hooks and a small component-level design language (tokens + Tailwind config) rather than only copied UI-kit code. There is clear composition and a global toast manager implemented as a lightweight in-memory store and listener pattern.
Evidence
victor-portfolio/components/ui/use-toast.ts: custom toast reducer, listeners and dispatch-based API (toast, useToast)
victor-portfolio/tailwind.config.js: custom design tokens, theme variants and safelist showing intent to maintain a design system
victor-portfolio/components/* and config/projects.ts: typed component/project metadata and use of TypeScript interfaces for component props/data
Responsive & Cross-browser
5/10
Works on all screens and browsers
Responsive and cross-browser concerns are addressed with Tailwind + custom CSS variables, theme modes, and media-query rules; the CSS contains careful container sizing and focus/hover states for different input widths. There is no evidence of container queries, RTL testing, or explicit feature-detection logic.
Evidence
victor-portfolio/globals.css: theme variables, multiple theme variants (dark, retro, cyberpunk...), media queries and responsive .card rules
victor-portfolio/tailwind.config.js: container config and extended theme tokens used for responsive layout
Performance Optimization
3/10
Speed of the interface
Some deliberate performance-minded choices are present (canvas/SVG chart smoothing, bounded in-memory buffers, caching layers), but there is no measurement artifacts (bundle analysis, Lighthouse numbers) or advanced code-splitting evidence in the human-authored frontend files.
Evidence
refer/polymarket-auto-trader/static/dashboard.js: priceHistory trimming (cap at 80 points) and Catmull-Rom smoothing (smoothPath) to keep rendering lightweight
MultiTradingByPolyterm/src/weather.js: fcCache and geoCache with TTLs to avoid repeated network work
Accessibility & Semantics
3/10
Usable for everyone
Basic accessibility and semantics are present (focus styles, keyboard-visible focus outlines, meaningful text fallbacks). However there is limited evidence of ARIA usage, keyboard handling for custom widgets, or automated a11y checks in CI.
Evidence
victor-portfolio/globals.css: :focus styles for .btn and focus-within rules for .card
refer/polymarket-auto-trader/static/dashboard.js: DOM-based controls with textual fallbacks and careful status/error banners
State Management & Data Flow
4/10
Managing data in the app
Async state discipline is pragmatic: debounced input lookups, polling cadence, caches with TTLs, and care for stale windows are visible. There is limited evidence of advanced server-state tooling (request cancellation tokens, optimistic updates with rollback, or a declarative cache invalidation strategy) in the human-authored frontend code.
Evidence
refer/polymarket-auto-trader/static/dashboard.js: input debounce (marketLookupTimer), periodic polling (setInterval refreshAll), and guarded retry messaging for transient boundary errors
MultiTradingByPolyterm/src/weather.js: fetchAllEvents pagination, geoCache and fcCache with TTL and fallback fetch logic (Open-Meteo → met.no)
UX & Visual Polish
6/10
Look and feel quality
UX and visual polish are strong for a small team/personal projects: theme variants, animated card interactions, informative empty/error states, chart area fills, and clear status pills contribute to a polished user experience. There are no pervasive skeleton loaders or detailed perceived-performance telemetry, but the UI provides good feedback on loading/error/empty cases.
Evidence
victor-portfolio/globals.css: multiple themes, well-crafted card hover/focus animations and blog prose styling
refer/polymarket-auto-trader/static/dashboard.js: empty-state messages, error banners, price-age stale indicator, and smooth chart drawing (area + path + target line)
Expertise
React• Middle
Modern Web Frameworks• Middle
Frontend Architecture & Build Tools• Middle
Industries
Financial Services• Middle
Technologies
TypeScript• Middle • 3y+
Zustand
Tailwind CSS
Next.js
Vite
Zod
Recommendations
- Own end-to-end data-driven dashboards and trading UIs that require reliable polling, charts, and error states (e.g., market-monitoring dashboards, ops panels).
- Build or extend Next.js product frontends that integrate with backend streams and server routes (SSG/SSR and API routes), using the existing design tokens and Tailwind system.
- Implement UX-critical features requiring precise time / timezone logic and numeric work (forecasting panels, event-driven notifications, or decision-support tooling).
- Collaborate on backend-integrated features where front-end state must mirror authoritative server state (settlement flows, manual buy/sell flows) to improve server-state discipline and optimistic update rollback.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior AI/ML Engineer
Confidence: Medium Generalist
A developer who builds algorithmic game-AI implementations and educational demos at a junior level. The strongest proven skill is implementing classical search algorithms - concretely seen in MancalaAI.minimax and MancalaAI.minimax_alpha_beta inside mancala-game-using-minimax.ipynb. The codebase does not show ML training pipelines, deployment, test automation, or production-grade MLOps practices.
Model Architecture & Training
2/10
How well models are designed and trained
No ML model training; implements classical game-search algorithms (minimax and alpha-beta) with a simple heuristic and class-based organization.
Evidence
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class MancalaAI.minimax
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class MancalaAI.minimax_alpha_beta
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class MancalaAI.evaluate_state
Data Pipeline & Feature Engineering
1/10
How data is prepared for models
Minimal state and input handling for the game board - basic initialization and move application, no feature engineering or dataset pipelines.
Evidence
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class Mancala.generate_board
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class Mancala.play
Experimentation & Evaluation
2/10
How results are measured and tested
Simple experiment/evaluation loop that collects wins/ties and average moves across multiple runs; results printed in the notebook but no experiment tracking or reproducible runs.
Evidence
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: function Mancala.match_analysis (prints wins/ties/average moves)
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: use of tqdm progress bar in match_analysis and printed timing/results
MLOps & Deployment
How models are shipped to production
Not evidenced in public code
Computational Efficiency
2/10
How efficiently computing resources are used
Basic algorithmic efficiency work - alpha-beta pruning is implemented and the notebook records elapsed times for different search depths, but there is no profiling, GPU usage, or advanced optimization.
Evidence
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: class MancalaAI.minimax_alpha_beta (alpha-beta pruning implementation)
Mancala-game-minimax-python/mancala-game-using-minimax.ipynb: notebook timing/results for depth 5 and depth 10 runs (elapsed time printed)
Research Depth & Innovation
Depth of research and new ideas
Not evidenced in public code
Technologies
Python• Senior • 6y+
C++• Senior • 6y+
MySQL
PostgreSQL• 5y+
Redis• 5y+
Rest API• 6y+
gRPC• 7y+
LangChain
Flask
SQLAlchemy
RabbitMQ• 5y+
Jupyter Notebook
FastAPI• 7y+
CI/CD• 5y+
Django• 5y+
PyTorch• 7y+
AWS• 9y+
Docker• 5y+
Apache Kafka
LLM• 7y+
RAG
Asyncio• 5y+
Celery• 5y+
PyTorch C++
Pydantic
STL
Alembic
Recommendations
- Develop small-scale game-AI components or educational algorithm demos that leverage minimax/alpha-beta and clear evaluation scripts.
- Harden the project with unit tests, reproducible notebook execution, and CI to make experimental results reproducible.
- Add performance profiling and iterative-deepening or move-ordering heuristics to improve search efficiency and demonstrate measurable gains.
- Prototype search-based agents or game-playing modules where deterministic, explainable decision logic is required.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Backend Developer
Confidence: Medium Generalist
A capable backend-focused generalist at a solid mid-level who independently implements domain-specific systems (trading/forecasting) with clear modular structure. The strongest proven skill is building numerics-driven trading logic and forecast pipelines, as evidenced by MultiTradingByPolyterm/src/weather.js (ensemble forecast, forecastDistribution, bandProb) and MultiTradingByPolyterm/refer/bot_v2.py (scan/update loop, calibration, state persistence). What is not evidenced is production-grade distributed architecture, database migrations, secrets rotation, or structured observability and long-running service orchestration.
API Design
4/10
How well APIs are designed
Practical API integration code with pagination and retry logic, but no explicit API versioning, idempotency or standardized error contract.
Evidence
MultiTradingByPolyterm/src/weather.js:fetchAllEvents implements manual pagination with offset/limit
MultiTradingByPolyterm/refer/bot_v2.py:get_ecmwf / get_hrrr use retries and timeouts around requests.get
Data Layer & Database
2/10
Working with databases
Lightweight local persistence and simple file-based storage; no migration history, transaction boundaries or tuned SQL.
Evidence
MultiTradingByPolyterm/refer/bot_v2.py:load_market / save_market write per-market JSON files under data/markets
MultiTradingByPolyterm/refer/bot_v2.py:load_state / save_state store runtime state in data/state.json
Scalability & Performance
4/10
Handling load and speed
Some caching and paging to limit work and network load, and careful model-ensemble computations, but no explicit queueing, sharding or measured performance optimization artifacts.
Evidence
MultiTradingByPolyterm/src/weather.js:fcCache and geoCache with FC_TTL_MS implement local caching and TTL
MultiTradingByPolyterm/src/weather.js:fetchAllEvents pages Gamma API in batches to avoid missing items
System Architecture
4/10
Overall system structure
Clear modular decomposition (forecasting, market parsing, execution loop) and runtime state handling; design is for a single-process trading service rather than distributed microservices.
Evidence
MultiTradingByPolyterm/refer/bot_v2.py:run_loop / scan_and_update separate main loop, scanning, monitor and resolution phases
MultiTradingByPolyterm/src/weather.js:forecastDistribution / discoverWeatherMarkets separate forecasting, parsing and market-discovery responsibilities
Security & Auth
2/10
Protecting data and access
Minimal boundary security practices visible - secrets read from config, some input parsing and basic guards present, but no token lifecycle, secrets management, or dependency auditing shown.
Evidence
MultiTradingByPolyterm/refer/bot_v2.py:get_actual_temp reads VC_KEY from config.json but no secret-management integration
MultiTradingByPolyterm/refer/bot_v2.py:parse_temp_range uses regex-based parsing to validate market questions
Reliability & Observability
4/10
Stability and monitoring
Reasonable operational hygiene: retries, timeouts, graceful KeyboardInterrupt handling and state saves; lacks structured observability, metrics and distributed resiliency patterns.
Evidence
MultiTradingByPolyterm/refer/bot_v2.py:requests calls use timeout tuples and retry loops; run_loop catches KeyboardInterrupt and saves state
MultiTradingByPolyterm/src/weather.js:openMeteoDaily falls back to met.no and sets a user-agent, showing defensive fallback design
Expertise
Python• Junior
Node.js• Junior
Technologies
Node JS• Middle
WebSockets
Requests
Dotenv
Recommendations
- Develop short-lived algorithmic trading services and single-process real-time pipelines that integrate multiple external HTTP feeds (forecasting, market data).
- Own forecasting and statistical modules that require careful numerical reasoning - e.g., ensemble forecasting, calibration, and probability-to-EV conversion.
- Implement medium-sized features around automation: market discovery, risk gating and book-reading adapters where single-process correctness and state persistence are primary.
- Improve production hardening: add structured logging/metrics, durable DB migrations, and secrets/credential management if the service moves to real money/live deployment.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: