George Michael
Full Stack Developer
7+ years exp
10+ projects
Java
SQL
C
C++
Python
TypeScript
JavaScript
Data Pipeline & Feature Engineering: 5/10
Model Architecture & Training: 4/10
Active 18 days ago
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
Full Stack Developer and technical team lead focused on enterprise CRM and AI-assisted systems, using FastAPI, React, and scalable REST APIs. Has experience delivering Spring Boot and Angular CRM features, maintaining production systems, and setting up Jenkins CI/CD pipelines. Also built Selenium-based automation test frameworks for CRM regression and smoke testing, improving reliability.
Technical skills
Java• Middle • 7y+ • 3 projects
SQL• Middle • 4y+ • 10+ projects
C• Middle • 3y+
C++• Middle • 3y+ • 5+ projects
Python• Senior
TypeScript• Senior • 10+ projects
JavaScript• Middle • 5+ projects
Python
HTTPX
Databases
MS SQL• 7y+ • 10+ projects
MySQL• 2 projects
PostgreSQL• 3 projects
AI/ML
NumPy
Pandas
Scikit-learn
OpenAI SDK
LLM
Transformers
NLP
RAG• 10+ projects
DevOps
Git
CI/CD• 7y+
Jenkins• 7y+
Rest API• 7y+
Docker
Analytics
Seaborn
QA
TestNG
Selenium• 3y+
Mobile
JUnit
IoT
MQTT
Frontend
Angular• 7y+
React.js
Design
Figma• 2 projects
Timeline
Team Lead / Lead Full Stack Developer
•
Lead
SKLADOFF 24
•
Full-Time
Led the architecture and development of an AI-driven CRM platform using FastAPI and React. Owned the end-to-end delivery cycle including backend/frontend implementation, deployment, testing, and team coordination. Built AI-assisted workflow automation with NLP features and analytics, and designed scalable REST APIs and asynchronous services. Implemented CI/CD and Docker-based infrastructure and established engineering practices through mentoring and code reviews.
Python
FastAPI
React.js
Rest API
NLP
CI/CD
Docker
Automation Testing Engineer
•
Middle
Zavod-IT
•
Full-Time
Built an automated testing framework for CRM systems using Selenium WebDriver with Java. Designed reusable automation components and applied the Page Object Model for stable test maintenance. Integrated automated suites into Jenkins-driven CI/CD pipelines for functional, regression, and smoke coverage. Improved test reliability by reducing flaky executions.
Selenium
Java
CI/CD
Jenkins
Project Engineer — Smart City / IoT
•
Middle
Honeywell
•
Full-Time
Contributed to an ACUD Smart City initiative in Egypt by supporting integration across IoT subsystems. Worked with international teams to align architecture and plan integration for smart parking, lighting, irrigation, and ICCC platforms. Supported backend integration and API coordination, including IoT communication layers. Produced UML diagrams and technical documentation for engineering reporting.
Full Stack Java Developer
•
Middle
EME International
•
Full-Time
Developed CRM functionality using Java Spring Boot and Angular. Implemented REST APIs and delivered frontend integrations, supporting end-to-end feature behavior. Managed production support by resolving incidents and optimizing system performance. Set up Jenkins-based CI/CD deployment workflows and collaborated with QA and business stakeholders to deliver enterprise solutions.
Javasince 2019
Spring Boot
Angular
Rest APIsince 2019
CI/CDsince 2019
Jenkinssince 2019
The British University in Egypt (BUE)
Bachelor's Degree •
ICS - informatics and computer science
Middle AI/ML Engineer
Confidence: Medium ML Engineer
A pragmatic ML/software engineer who builds applied systems - conversational bots and Kaggle-style tabular ML pipelines - at a solid middle level. The strongest proven skill is end-to-end applied ML and modeling, demonstrated by the house_prices_prediction_kaggle/main.py pipeline (data cleaning, feature engineering, cross-validation, multiple models and ensembling). Missing from public code are automated tests, experiment tracking, deployment observability and advanced LLM engineering (RAG, guardrails, streaming inference).
Model Architecture & Training
4/10
How well models are designed and trained
Solid application of classical ML models and ensembling for tabular prediction, but no custom architectures or advanced training infrastructure.
Evidence
house_prices_prediction_kaggle/main.py: training blocks for Ridge, Lasso, ElasticNet, RandomForestRegressor, GradientBoostingRegressor and ensemble prediction
house_prices_prediction_kaggle/main.py: use of cross_val_score and KFold for model validation
Data Pipeline & Feature Engineering
5/10
How data is prepared for models
Comprehensive tabular data preprocessing and feature engineering (imputation, new features, skew handling, one-hot encoding) appropriate for Kaggle-style workflows.
Evidence
house_prices_prediction_kaggle/main.py: missing_data function and multiple imputation rules (none_features, zero_features, group median LotFrontage)
house_prices_prediction_kaggle/main.py: engineered features (TotalSF, TotalBath, HouseAge, HasGarage, etc.) and categorical encoding via pd.get_dummies
Experimentation & Evaluation
3/10
How results are measured and tested
Basic experiment/evaluation practices are present (CV, KFold, RMSE reporting) but no experiment tracking, ablation studies, or reproducibility tooling.
Evidence
house_prices_prediction_kaggle/main.py: cross_val_score with KFold and RMSE calculations
house_prices_prediction_kaggle/main.py: printing and saving diagnostic plots (saleprice_distribution.png, feature_importance.png)
MLOps & Deployment
3/10
How models are shipped to production
Production-focused FastAPI webhook with secret verification and webhook management is implemented, but there is limited deployment/observability/MLOps automation.
Evidence
LAB/app/main.py: /webhook endpoint and verify_telegram_secret implementation (HMAC compare_digest)
LAB/app/telegram.py: set_webhook and delete_webhook functions using Telegram API
LAB/scripts/set_webhook.py: script placeholder to set/delete webhook (integration automation)
Computational Efficiency
2/10
How efficiently computing resources are used
Some practical efficiency choices (async httpx usage, RandomForest n_jobs) are present, but no GPU/quantization/distributed or profiling work.
Evidence
LAB/app/telegram.py and LAB/app/openai_client.py: use of httpx.AsyncClient for non-blocking I/O
house_prices_prediction_kaggle/main.py: RandomForestRegressor configured with n_jobs=-1
Research Depth & Innovation
1/10
Depth of research and new ideas
No evidence of original research, custom model architectures, or reproduced papers - mostly applied, conventional techniques.
Evidence
house_prices_prediction_kaggle/main.py: standard modelling pipeline and ensemble approach without novel algorithms or custom layers
Expertise
Conversational AI & Chatbots• Middle
LLM• Middle
Industries
Artificial Intelligence• Middle
Technologies
NLP
LLM
Recommendations
- Develop webhook-backed LLM chatbots with Redis-backed per-user state and retries/backoff for API calls (replace in-memory history for production).
- Add CI, unit tests and lightweight experiment tracking (MLflow or W&B) to make ML experiments reproducible and auditable.
- Harden OpenAI integration with rate-limit handling, exponential backoff, structured error handling and secrets management (vault or cloud secret manager).
- Extend tabular pipeline to include hyperparameter tuning (Ray Tune / Optuna) and try gradient boosting frameworks (XGBoost/LightGBM) with proper validation pipelines.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: Medium ML Practitioner
A machine-learning practitioner at an early senior level (tier_score 3.5) who independently implements end-to-end models and lightweight async production services. The strongest proven skill is applied predictive modeling and feature engineering, demonstrated by the house_prices_prediction_kaggle/main.py pipeline with detailed imputation, feature creation, scaling, KFold cross-validation and an ensemble submission flow. There is limited evidence of robust MLOps, automated hyperparameter tuning, unit/ integration tests or production-level data-versioning and CI/CD practices.
Statistical Rigor
4/10
Correct use of statistics
Some sound statistical practices (skewness checks, log transforms, KFold CV and basic classification reports) but no formal inference, uncertainty quantification, multiple-comparison controls, causal analysis or principled hypothesis testing.
Evidence
house_prices_prediction_kaggle/main.py: skewness/kurtosis reporting and log1p transformation; KFold + cross_val_score usage
NLP_Text_categorization/Sol2_trainingModel.py: use of classification_report and accuracy_score
Data Wrangling & Cleaning
6/10
Preparing and cleaning data
Clear, deliberate missing-value handling and multiple imputation strategies, type-aware fills, group-based median imputation and a range of engineered features - good practical data-wrangling for tabular problems.
Evidence
house_prices_prediction_kaggle/main.py: missing_data function and multiple imputation blocks (none_features, zero_features, group median LotFrontage)
house_prices_prediction_kaggle/main.py: feature engineering (TotalSF, TotalBath, HasGarage, HouseAge, etc.)
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Includes EDA plots, saved figures and printed summaries (distribution, skewness, feature importances) but limited written interpretation per figure and no structured storytelling tying findings to modeling choices.
Evidence
house_prices_prediction_kaggle/main.py: saleprice_distribution.png generation and printed skew/kurtosis outputs
house_prices_prediction_kaggle/main.py: feature_importance.png saved and printed top-20 importances
Predictive Modeling
5/10
Building models that predict
Trains multiple standard models with scaling, cross-validated RMSE and a simple weighted ensemble - demonstrates baseline-first discipline and evaluation but lacks automated hyperparameter tuning, calibration, deeper error analysis or robustness checks.
Evidence
house_prices_prediction_kaggle/main.py: training Ridge, Lasso, ElasticNet, RandomForest, GradientBoosting with cross_val_score and ensemble weighting
NLP_Text_categorization/Sol2_trainingModel.py: TF-IDF + LogisticRegression pipeline with train/test split and evaluation
Business Insight & Impact
2/10
Turning analysis into business value
Minimal evidence of business framing or impact analysis - no discussion of business metrics, error costs, or how model outputs map to decisions or KPIs.
Evidence
house_prices_prediction_kaggle/main.py: 'Next steps' are technical (Kaggle upload, tuning) with no business-metric discussion
Reproducibility & Notebook Hygiene
3/10
Clean, repeatable analysis
Some reproducibility signals (random_state seeds, a requirements.txt and pydantic-based settings) but missing pinned environment files, data/versioning (DVC), CI, test suites or notebook execution metadata for full reproducibility.
Evidence
LAB/requirements.txt: lists FastAPI, uvicorn and other runtime deps
house_prices_prediction_kaggle/main.py: KFold(random_state=42) and model random_state parameters
LAB/app/config.py: Settings class using pydantic SettingsConfigDict for env-file config
Expertise
Analytics• Middle
Technologies
Scikit-learn
Seaborn
OpenAI SDK
Transformers
Pandas
NumPy
Recommendations
- Develop production Telegram bots and lightweight async APIs that integrate LLMs and external APIs (use the LAB FastAPI/OpenAI code path).
- Build end-to-end supervised ML pipelines for tabular prediction problems, focusing on feature engineering, cross-validation and ensembling.
- Prototype NLP classification solutions using TF-IDF or transformers for practical labeling tasks.
- Strengthen reproducibility and production readiness by adding CI, tests, data/versioning (DVC) and hyperparameter tuning workflows (Optuna/Hyperopt).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Backend Developer
Confidence: Medium API Engineer
A pragmatic backend engineer at a middle level who builds async webhook integrations and small API services in Python. The strongest proven skill is integrating webhook-driven chat workflows and LLM APIs, evidenced by the FastAPI /webhook handler (app/main.py) and the OpenAI client integration (app/openai_client.py). There is limited evidence of persistent-data design, testing, CI, advanced scaling (queues/caching), or observability beyond basic logging.
API Design
4/10
How well APIs are designed
Concrete webhook and API handling for Telegram with explicit secret verification and consistent 200 response semantics; lacks advanced API versioning, idempotency handling, or pagination patterns.
Data Layer & Database
1/10
Working with databases
No persistent datastore, migrations, or transaction handling - only an in-memory per-user history and a MAX_HISTORY setting.
Scalability & Performance
3/10
Handling load and speed
Async design with httpx timeouts and message chunking shows awareness of I/O and request limits, but there is no caching, queuing, connection pool tuning, or measured scaling work.
System Architecture
3/10
Overall system structure
Reasonable small-service modular decomposition (main, bot, telegram, openai_client, config, history) appropriate for a single-process webhook service; no evidence of multi-service orchestration or advanced service decomposition decisions.
Security & Auth
4/10
Protecting data and access
Conscious secret handling and header verification (hmac.compare_digest), environment-backed settings for secrets, and guarded webhook setup - good basic auth/security hygiene for webhook integrations.
Reliability & Observability
4/10
Stability and monitoring
Basic reliability and observability patterns present - logging, exception capture, lifespan hooks and timeouts - but no structured tracing, metrics, retries with backoff or graceful shutdown sequencing beyond the simple lifespan hook.
Expertise
Backend AI & LLM• Junior
Python• Junior
Industries
Artificial Intelligence• Middle
Technologies
HTTPX
Recommendations
- Own webhook-driven AI/chat integrations and build production hardening - add Redis-backed session/history store, rate limiting, and idempotency handling for webhooks.
- Develop async HTTP APIs and integrations (FastAPI + httpx) for small-to-medium services, focusing on retry/backoff, connection pooling, and structured logging/metrics.
- Add CI/tests and implement observability (metrics, tracing, alerts) and a simple deployment pipeline (Docker/Uvicorn + basic process manager) to make the service production-ready.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: