Raafat Basheer
Technical Lead
6+ years exp
5+ years ML exp
Management: Under a year
TypeScript
Node JS
JavaScript
Swift
Python
Java
Kotlin
Objective-C
Data Wrangling & Cleaning: 6/10
Exploratory Analysis & Visualization: 5/10
Predictive Modeling: 5/10
Invite to interview
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
ML practitioner (senior-level) focused on production-ready model training and inference services for forecasting and image-based classification. The strongest proven skill is engineering end-to-end ML systems as shown by the custom seq2seq forecaster, training/evaluation pipeline, and the FastAPI inference service with comprehensive tests. There is little public evidence of large-scale distributed data engineering, managed MLOps (DVC/Kubeflow), or cloud-native production deployments and monitoring pipelines.
Phone
Technical skills
Languages
8
TypeScript
Node JS
JavaScript
Swift
Python
Java
Kotlin
Objective-C
Frontend
5
React.js
Angular
Vue.js
Next.js
Sass
DevOps
4
Docker
AWS
Rest API
CI/CD
AI/ML
7
Copilot
LLM
Machine Learning
NumPy
Pandas
Scikit-learn
Pillow
Other
30
Mockito
JUnit
Spring Boot
PCI DSS
Click
MongoDB
Matplotlib
Bcrypt
Axios
Express
Mongoose
Nest.JS
FastAPI
Flask
Kubernetes
PyTorch
TensorFlow
Transformers
MySQL
Cypress
Prompt Engineering
Keras
GitHub
Deep Learning
Figma
Git
Dotenv
Jenkins
AI Agents
RAG
Timeline
Co-Founder & Technical Lead
•
Head+
Startup in formation
•
Full-Time
Co-founded an early-stage startup and led the technical architecture for a planned SaaS product for web and mobile. Designed an AI-powered system based on Transformer and LLM approaches and evaluated alternative model architectures to choose a technical direction. Owned key early technical decisions, including stack selection and scalability planning for future growth.
Transformers
LLM
Full-Stack Developer
•
Middle
Adorsys GmbH
•
Full-Time
Developed and maintained security-critical FinTech applications with a focus on regulatory compliance, data protection, and scalability. Built frontend modules with TypeScript, Angular, and Vue.js and backend services using Node.js, including REST APIs for payment and transaction logic. Implemented OAuth2/OpenID Connect authentication, improved CI/CD with Docker-based pipelines, and introduced automated test suites to prevent release issues.
TypeScript
Angular
Vue.js
Node JS
Docker
CI/CD
JUnit
Mockito
Copilot
PCI DSS
Rest API
Scrum
Software Developer
•
Middle
yourgix.com GmbH
•
Full-Time
Built a music event platform supporting ticket sales and live streaming. Developed interactive, performant user interfaces with React and Next.js, and implemented backend REST APIs for event, user, and streaming management with Node.js. Used AWS services for video storage and live-streaming workflows, and standardized delivery via Docker-based deployment.
React.js
Next.js
Node JS
Docker
AWS
TypeScript
Tutor & Development Assistant
•
Middle
DCI Career Institute
•
Full-Time
Supported students through debugging, code reviews, and guided project work in modern web development. Assisted with development of JavaScript and React-based features and helped connect backend work with Node.js. Collaborated with instructors to apply best practices and improve learning outcomes across student projects.
JavaScript
React.js
Node JS
University of Aleppo
Bachelor's Degree •
Business Administration
Senior Data Scientist
Confidence: Medium ML Practitioner
ML practitioner (senior-level) focused on production-ready model training and inference services for forecasting and image-based classification. The strongest proven skill is engineering end-to-end ML systems as shown by the custom seq2seq forecaster, training/evaluation pipeline, and the FastAPI inference service with comprehensive tests. There is little public evidence of large-scale distributed data engineering, managed MLOps (DVC/Kubeflow), or cloud-native production deployments and monitoring pipelines.
Statistical Rigor
4/10
Correct use of statistics
Solid forecast evaluation and custom loss design are present, but formal statistical testing, uncertainty quantification beyond point metrics, and causal analysis are limited.
Evidence
electricity-forecast/src/evaluate.py: horizon-bucket MAE/RMSE/MAPE evaluation
FreshSight_AI/backend/scripts/train.py: class_weights and validation tracking for classifier
Data Wrangling & Cleaning
6/10
Preparing and cleaning data
Clear attention to leakage prevention and reproducible preprocessing; scaler persistence and windowed dataset splitting show good data-wrangling discipline.
Evidence
electricity-forecast/src/train.py: prepare_data uses chronological_boundaries and fit_scaler; saves scaler to checkpoint
electricity-forecast/src/train.py: WindowedDataset usage to construct leakage-free train/val/test sets
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Practical, production-ready visual reporting (interactive Plotly HTML) that computes and surfaces forecast metrics, but limited exploratory narrative or deeper EDA artifacts.
Evidence
electricity-forecast/src/report.py: builds interactive Plotly report with history, forecast, residuals and summary table
Predictive Modeling
5/10
Building models that predict
Custom seq2seq forecasting model, scheduled sampling (teacher forcing ratio), explicit training loop and evaluation pipeline demonstrate competent predictive modeling engineering.
Evidence
electricity-forecast/src/models/seq2seq.py: custom GRU/LSTM encoder-decoder with scheduled-sampling logic
electricity-forecast/src/train.py: full training loop, weighted_loss, LR scheduler, early stopping and checkpointing
FreshSight_AI/backend/scripts/train.py: efficientnet transfer-learning training script with class reweighting
Business Insight & Impact
3/10
Turning analysis into business value
Work connects predictions to product features (quality score, recommendations, forecast reports), but explicit business-metric analysis, cost-of-error reasoning, or stakeholder impact modeling is minimal.
Evidence
FreshSight_AI/backend/app/services/inspection_service.py: computes quality_score and recommendation from model outputs
electricity-forecast/src/report.py: generates forecast metrics table (MAE/RMSE/MAPE) for stakeholders
Reproducibility & Notebook Hygiene
5/10
Clean, repeatable analysis
Good reproducibility practices: requirements, CLI, seeding, unit tests, dependency overrides and saved artifacts; no evidence of DVC/data versioning or full CI/CD manifests for model lifecycle.
Evidence
electricity-forecast/requirements.txt and FreshSight_AI/backend/requirements.txt: pinned dependencies
electricity-forecast/src/train.py: seed_everything for deterministic runs
FreshSight_AI/backend/tests/conftest.py and multiple tests: comprehensive unit/test fixtures and dependency overrides
Expertise
Data Science• Senior
Industries
Energy & Utilities• Middle
Food & Beverages• Middle
Technologies
Deep Learning
Python• since 2026 • Senior
MongoDB
Scikit-learn• since 2026
Matplotlib
Pandas• since 2026
NumPy• since 2026
Click
Pillow
Machine Learning• since 2025
Recommendations
- Lead development of production ML services that include model training, evaluation, and a REST inference API (PyTorch + FastAPI).
- Implement time-series forecasting systems and consumer-facing forecast reporting dashboards with interactive visualizations.
- Build image-classification training workflows and lightweight model serving for mobile/web products, including transfer-learning pipelines.
- Contribute to unit-tested model logic and validation layers for safety-critical ML features (input validation, confidence-thresholding, explainable scoring).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle AI/ML Engineer
Confidence: High ML Engineer
ML engineer focused on practical applied ML and production-capable services, operating at a solid middle level with strengths in end-to-end model pipelines and service integration. The strongest proven skill is building reproducible training and evaluation pipelines, including a custom autoregressive seq2seq forecaster and held-out horizon evaluation as shown in electricity-forecast/src/train.py and src/evaluate.py. There is limited evidence of large-scale distributed training, quantization/efficiency engineering, or platform-level MLOps (Kubernetes/monitoring/CI deployment pipelines).
Model Architecture & Training
6/10
How well models are designed and trained
Solid, practical model building and custom training loops with configurable losses, scheduled sampling (teacher forcing), gradient clipping, and checkpointing; typical of an ML engineer building production-capable models rather than research-grade novel architectures.
Evidence
electricity-forecast/src/train.py:train (seed_everything, weighted_loss, training loop, scheduler, checkpoint save)
electricity-forecast/src/models/seq2seq.py:Seq2SeqForecaster (autoregressive encoder-decoder with teacher forcing)
FreshSight_AI/backend/scripts/train.py:build_model and run_epoch (transfer learning fine-tune of EfficientNet)
Data Pipeline & Feature Engineering
6/10
How data is prepared for models
Clear, leakage-aware data preparation and scaling, windowed dataset construction, and feature creation for time-series forecasting; also standard image preprocessing for CV training.
Evidence
electricity-forecast/src/train.py:prepare_data (fits scaler, creates WindowedDataset splits)
electricity-forecast/src/predict.py:predict (uses calendar_features and load_scaler)
FreshSight_AI/backend/app/model/preprocessing.py:build_transform (image preprocessing used by CV training script)
Experimentation & Evaluation
6/10
How results are measured and tested
Reproducible experiments and evaluation pipeline with held-out test evaluation, horizon-bucket metrics, Plotly reporting and CSV/JSON logging; tests exercise CLI and model/test flows.
Evidence
electricity-forecast/src/evaluate.py:evaluate (horizon-bucket MAE/RMSE/MAPE and JSON output)
electricity-forecast/src/report.py:build_report (interactive Plotly report generation and metric table)
electricity-forecast/tests/test_cli_smoke.py (CLI smoke test covering predict/report paths)
MLOps & Deployment
4/10
How models are shipped to production
Production-aware service design for inference with FastAPI, dependency injection, lifespan/startup checks and error handling, but without evidence of full infra (Kubernetes, CI/CD pipelines, or monitoring) for large-scale deployment.
Evidence
FreshSight_AI/backend/app/main.py:create_app and lifespan (startup checks, predictor build, AppError handlers)
FreshSight_AI/backend/app/dependencies.py:get_predictor and related DI helpers (predictor/repo/storage wiring)
Computational Efficiency
3/10
How efficiently computing resources are used
Some efficiency awareness (device selection, num_workers, gradient clipping, scheduler) and standard transfer-learning choices, but no measured GPU/throughput profiling, quantization, or advanced efficiency engineering.
Evidence
FreshSight_AI/backend/scripts/train.py:main (device selection and DataLoader num_workers)
electricity-forecast/src/train.py:training loop (clip_grad_norm_, scheduler usage, batch sizing)
Research Depth & Innovation
2/10
Depth of research and new ideas
Applied implementations of common architectures and techniques (encoder-decoder forecasting, transfer learning) but no evidence of novel research contributions, custom layers, or reproduced SOTA experiments.
Evidence
electricity-forecast/src/models/seq2seq.py:Seq2SeqForecaster (standard GRU/LSTM encoder-decoder)
FreshSight_AI/backend/scripts/train.py (EfficientNet transfer learning training loop)
Expertise
Computer Vision & Image Analysis• Middle
Industries
Energy & Utilities• Middle
Food & Beverages• Middle
Technologies
MySQL
Copilot• since 2021
Prompt Engineering• since 2025
AI Agents
CI/CD• since 2021
Jenkins
Transformers• since 2026
TensorFlow• since 2026
Keras• since 2026
Git
PyTorch• since 2026
AWS• since 2021
Docker• since 2021
Kubernetes
LLM• since 2025
RAG
GitHub
Recommendations
- Lead development of end-to-end ML features such as time-series forecasting products or CV classifiers where you own data prep, training loop, evaluation and inference serving (use the electricity-forecast and FreshSight AI patterns).
- Productionize model serving and monitoring for inference services (add CI/CD, containerization, observability, drift detection and structured metrics).
- Expand efficiency engineering by adding GPU profiling, batching/throughput tests, and model optimizations (quantization or TorchScript) for lower-latency inference.
- Build standardized experiment tracking (W&B/MLflow) and structured ablation runs to strengthen reproducibility and model selection.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Backend Developer
Confidence: High API Engineer
Full-stack web developer (Junior) focused on simple web applications and authentication flows with React/Next.js frontends and a Node/Express backend. The strongest proven skill is building basic authentication and user persistence using Express and Mongoose as seen in react-hook/backend/models/userModel.js and react-hook/backend/controllers/userController.js. There is little evidence of production-grade concerns such as migration history, automated tests, observability, or robust security hardening in public code.
API Design
2/10
How well APIs are designed
Basic REST endpoints implemented but missing API-level design practices such as versioning, idempotency, consistent error contract, and structured input validation; frontend calls backend directly with simple axios requests.
Evidence
react-hook/backend/controllers/userController.js: login/register/logout handlers using jwt and res.cookie
react-hook/backend/routes/userRouter.js: presence of router file for user endpoints
react-hook/src/components/Login.js and react-hook/src/components/Register.js: axios POST calls to backend endpoints
Data Layer & Database
3/10
Working with databases
Uses Mongoose schema with uniqueness constraints and bcrypt hashing, but query/async mistakes and no migration history, transactions, or tuned queries are present.
Evidence
react-hook/backend/models/userModel.js: Mongoose schema, bcrypt.hashSync in registerUser, loginUser implementation
react-hook/backend/lib/db.js: mongoose import and DB connection setup
Scalability & Performance
2/10
Handling load and speed
Almost no backend scalability features such as caching, queuing, or rate limiting; frontend config includes asset caching headers and compression for static assets.
Evidence
automation-landing-page/next.config.ts: Cache-Control headers and compress=true
react-hook/backend/package.json: typical small backend deps with no caching or queue libraries
System Architecture
2/10
Overall system structure
Monorepo with a clear frontend/backend split but otherwise a single-service design with limited module/service decomposition or inter-service contracts.
Evidence
react-hook/backend/app.js: top-level express app import and wiring
react-hook/package.json: separate frontend project with React and axios showing frontend-backend separation
Security & Auth
3/10
Protecting data and access
Implements bcrypt and JWT-based cookie auth which is a positive signal, but token handling and error handling are weak (no expiry, whole user embedded in token), and input validation is not evidenced.
Evidence
react-hook/backend/models/userModel.js: bcrypt used for hashing passwords
react-hook/backend/controllers/userController.js: jwt.sign usage and setting authCookie
react-hook/backend/middleware/verifyToken.js: presence of token verification middleware
Reliability & Observability
1/10
Stability and monitoring
Minimal reliability and observability practices; basic try/catch present but no structured logging, metrics, retries with backoff, timeouts, or graceful shutdown patterns.
Evidence
react-hook/backend/controllers/userController.js: try/catch blocks returning raw errors
automation-landing-page/next.config.ts: security headers present but no observability/metrics configuration
Expertise
Node.js• Junior
Microservices & API Architecture• Junior
Databases & Vector Storage• Junior
Technologies
Java
Kotlin
Rest API• since 2021
Flask
Express
Spring Boot• since 2023
FastAPI
Bcrypt
Mockito• since 2021
Mongoose
Axios
Dotenv
Recommendations
- Harden backend authentication: add token expiry, sign minimal claims, avoid embedding entire user objects in JWTs, and add input validation middleware for endpoints.
- Fix async bugs and improve data access: await Mongoose queries, add error classes, and introduce migration/versioning tools for schema evolution.
- Add basic observability and resilience: structured logging, request timeouts, retries with exponential backoff where appropriate, and simple metrics/health endpoints.
- Write tests and CI: add unit and integration tests for auth flows and database interactions and run them in CI to catch regressions early.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: