Zohaer Al Mahatab
Artificial Intelligence Specialist
Node JS
JavaScript
TypeScript
Python
Data Pipeline & Feature Engineering: 5/10
Model Architecture & Training: 4/10
Active 1 day ago
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A middle-level ML engineer specializing in end-to-end, data-driven market discovery and related scientific analysis workflows. The strongest proven skill is modular feature engineering and temporally-safe label construction as demonstrated by features/feature_pipeline.py and models/trainer.py. There is little public evidence of experiment tracking, production model serving, GPU/distributed training, or systematic backtesting and monitoring.
Technical skills
Node JS
JavaScript
TypeScript
Python• Middle
Node JS
Express
Prisma
Python
Requests
Pydantic
Databases
MySQL
PostgreSQL
AI/ML
NumPy
Pandas
RAG
Scikit-learn
XGBoost
Jupyter Notebook
Streamlit
huggingface_hub
NLTK
Frontend
Next.js
React.js
Three.JS
React Three Fiber
Framer Motion
Tailwind CSS
DevOps
Docker
Analytics
Matplotlib
Mobile
JUnit
QA
Pytest
Timeline
Data Science and Machine Learning Intern
•
Junior
Valkyrie
•
Internship
Built a real-time SLAM navigation system (ROS2, LIDAR, Jetson, OpenCV) enabling a drone to autonomously map GPS-denied indoor environments, detect ArUco targets via computer vision, and land within 60 seconds using PnP-based visual servoing.
Built a live NFL win-probability model (XGBoost on play-by-play data) and measured a 20s Polymarket live-feed lead over broadcast, isolating an 18s in-game trading edge window across Kalshi/Polymarket sports markets.
Contributed to an automated scoring workflow for real-estate buy-box fit and investment projection across a Texas Triangle context. Focused on developing and validating components for downstream decision-support.
Python
Scikit-learn
ROS2
JavaScript
OpenCV
AWS
XGBoost
St. Edward's University
Bachelor's Degree •
Computer Science (Artificial Intelligence and Machine Learning concentration)
Data Scientist
•
Middle
St. Edward’s University
•
Full-Time
Handled large datasets for analysis and model/data preparation tasks. Worked on improving and correcting existing radiosonde data records to increase data quality for subsequent use. Supported research-oriented data processing and validation workflows.
Middle AI/ML Engineer
Confidence: High ML Engineer
A middle-level ML engineer specializing in end-to-end, data-driven market discovery and related scientific analysis workflows. The strongest proven skill is modular feature engineering and temporally-safe label construction as demonstrated by features/feature_pipeline.py and models/trainer.py. There is little public evidence of experiment tracking, production model serving, GPU/distributed training, or systematic backtesting and monitoring.
Model Architecture & Training
4/10
How well models are designed and trained
Solid applied model work using XGBoost with a careful chronological train/test split and explicit label construction to avoid lookahead, but no custom architectures, advanced hyperparameter search, or rigorous cross-validation.
Data Pipeline & Feature Engineering
5/10
How data is prepared for models
Clear modular feature engineering and validation pipeline with separate price/volume/sector/sentiment feature modules and a feature matrix builder; includes validation checks and ETF/sector mappings.
Experimentation & Evaluation
2/10
How results are measured and tested
Basic evaluation via a held-out chronological split and a printed accuracy score plus lightweight tests; lacks experiment tracking, multi-metric evaluation, ablations or reproducible experiment config.
MLOps & Deployment
3/10
How models are shipped to production
Some operational work is present: model artifact persistence and a scheduled daily runner pipeline for end-to-end runs, but no production serving, monitoring, versioning, or deployment automation shown.
Evidence
priom-mahatab/stock-radar/models/trainer.py: joblib.dump(model, MODEL_URL) persists model artifacts
priom-mahatab/stock-radar/pipeline/scheduler.py: imports APScheduler and wires a pipeline runner; priom-mahatab/stock-radar/pipeline/daily_runner.py orchestrates ingestion, features, prediction and reporting
Computational Efficiency
1/10
How efficiently computing resources are used
No evidence of GPU optimization, quantization, distributed training, or profiling; training uses default XGBoost settings without explicit efficiency work.
Research Depth & Innovation
3/10
Depth of research and new ideas
Moderate research-oriented experimentation in a satellite notebook showing clustering analysis (KMeans, HDBSCAN, silhouette) for lidar data, but research artifacts are exploratory notebooks rather than production-grade paper implementations.
Verified artifacts
Expertise
Finance & FinTech AI• Middle
Scientific AI (Life Sciences, Physics, Climate)• Middle
Industries
Financial Services• Middle
Science & Engineering• Middle
Technologies
MySQL
PostgreSQL
RAG
Recommendations
- Stabilize and test the end-to-end pipeline by adding unit tests and CI for feature functions and trainer/predictor paths so model regressions are caught early.
- Introduce experiment tracking (MLflow or Weights & Biases) and structured eval scripts to record metrics, baselines and hyperparameter sweeps for reproducibility.
- Add a lightweight model serving layer and model versioning (e.g., a simple REST endpoint or BentoML) plus monitoring for data drift and prediction quality.
- Consolidate and minimize requirements.txt to remove unused packages and add dependency security scanning and pinned versions for deployment safety.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: High ML Practitioner
A practical ML practitioner (middle level) focused on building feature pipelines and end-to-end model training for market and remote-sensing data. The strongest proven skill is building end-to-end feature pipelines and training workflows, as shown by features/feature_pipeline.py and models/trainer.py that implement lookahead-safe labeling, chronological splits, and model persistence. What is not evidenced is production-grade deployment, robust model monitoring, extensive hyperparameter search, or formal experiment tracking and CI/CD pipelines.
Statistical Rigor
3/10
Correct use of statistics
Basic statistical rigor is present (chronological split, clustering silhouette checks) but there is limited uncertainty quantification, hypothesis testing, or deeper error/robustness analysis; evaluation is mainly single-score reporting.
Data Wrangling & Cleaning
5/10
Preparing and cleaning data
Shows solid data-wrangling practices: per-ticker validation, explicit handling of lookahead, masking/removal of invalid lidar values, and thoughtful flattening of arrays; some provenance via a cache exists but more formal validation/monitoring would raise this score.
Evidence
priom-mahatab/stock-radar/features/feature_pipeline.py: uses validate_price_df and per-ticker feature construction
priom-mahatab/stock-radar/models/trainer.py: excludes final 5 dates for forward-labeling to avoid leakage and uses sector mapping
ARM-Synergy/ndim-auto-cluster-lidar/notebooks/hsrl_analysis.ipynb: remove_negatives, valid_mask and flattening to exclude NaNs before clustering
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Exploratory analysis and visualization are purposeful in the lidar notebook with written interpretations and cluster-selection plots; the stock project has limited in-code EDA and fewer documented interpretations.
Predictive Modeling
3/10
Building models that predict
Predictive modeling is practiced (feature pipeline, XGBoost model training, persistence) with careful chronological splitting, but lacks baselines, cross-validation or in-depth error analysis, calibration, and hyperparameter tuning.
Business Insight & Impact
2/10
Turning analysis into business value
Some product thinking is visible (report formatting, threshold constants for signals), but there is little explicit linking of model metrics to business KPIs or cost-sensitive error analysis (FP vs FN tradeoffs).
Reproducibility & Notebook Hygiene
5/10
Clean, repeatable analysis
Reproducibility basics are present: pinned requirements.txt, seeded models, tests for ingestion/pathways, and saved artifacts; however, there is no evidence of CI, data/versioning (DVC), or deployment pipelines.
Expertise
Analytics• Middle
Industries
Financial Services• Middle
Science & Engineering• Middle
Technologies
Python• Middle
Jupyter Notebook
XGBoost
Scikit-learn
huggingface_hub
Pandas
NumPy
NLTK
Streamlit
Pydantic
Requests
Recommendations
- Productionize model lifecycle: add CI runs, automated training/validation jobs, and model registry or artifact versioning (promote joblib persistence to a registry).
- Strengthen evaluation: implement baselines, cross-validation or time-series CV, calibration checks, and more detailed error analysis (confusion matrices, per-sector metrics).
- Improve data reliability: add structured data validation rules, monitoring/alerts for data drift, and dataset provenance/versioning (DVC or similar).
- Build backtesting and business metrics: add historical backtests and explicit FP/FN cost modeling to connect model outputs to measurable business impact.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Frontend Developer
Confidence: Medium UI Engineer
A frontend UI engineer at a junior level focusing on interactive geospatial visualizations and portfolio-style work. The strongest proven skill is building map visualizations using Leaflet/Folium as shown by the us-electricity-cost-heatmap index.html map and heat-layer data. There is little evidence of component architecture, async state management, testing, accessibility, or measured performance optimization in public code.
UI Component Architecture
2/10
How interface parts are built
Minimal component architecture; mostly static HTML output and no evidence of modular component design or custom reusable UI primitives.
Responsive & Cross-browser
3/10
Works on all screens and browsers
Basic responsive layout via viewport meta and container styles for the map, but no advanced responsive techniques or explicit RTL/i18n readiness.
Performance Optimization
1/10
Speed of the interface
No measurable performance work, only default Leaflet tile and heat layers; no code-splitting, bundle analysis, virtualization or proven optimizations.
Accessibility & Semantics
1/10
Usable for everyone
Little to no accessibility work visible; map output lacks ARIA landmarks and there is no evidence of keyboard focus management for custom controls.
State Management & Data Flow
1/10
Managing data in the app
No async state discipline or client-side state management; data is embedded statically and there is no request cancellation, optimistic updates, or state machines.
UX & Visual Polish
3/10
Look and feel quality
Functional UX for a map visualization with a fixed-position legend and map controls; visually usable but lacks richer UX patterns like skeletons, progressive loading or undo.
Expertise
Modern Web Frameworks• Junior
Industries
Energy & Utilities• Junior
Technologies
JavaScript
TypeScript
Node JS
Tailwind CSS
Next.js
Express
Three.JS
Prisma
Docker
React.js
Framer Motion
React Three Fiber
Recommendations
- Develop interactive data-visualization dashboards and mapping UIs (Leaflet/Mapbox/React wrappers) where you implement componentized map controls, accessibility, and stateful filtering.
- Build modern React/Next.js pages that convert generated static HTML into reusable components and add client-side state (cancellation, race handling, optimistic UI) and tests.
- Improve accessibility and keyboard support for custom map controls and legend widgets, and add ARIA roles and focus management.
- Work on performance improvements that are measurable: lazy-load heavy libraries, code-splitting, and instrument LCP/INP or bundle-analyzer before/after.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: