Muhammad Shoaib
Data Scientist
6+ years exp
Management: 6-10 years
Python
C++
SQL
Java
C
MATLAB
Fortran
Model Architecture & Training: 6/10
Research Depth & Innovation: 6/10
Data Pipeline & Feature Engineering: 5/10
Active 13 days ago
Invite to interview
Download CVCV
Overview
Technical skills
Projects
Timeline
Roles
Overview
Quantitative finance researcher and applied statistics engineer (senior level) with demonstrated strengths in careful econometric inference and reproducible benchmarking for portfolio performance; the strongest proven skill is rigorous cross-sectional and time-series inference implemented in analysis/latent_performance.py and the HalfNormal SFA implementation in sfa/sfa_halfnormal.py. The codebase shows comprehensive unit and integration tests, deterministic report generation, and a deployable Streamlit/FastAPI MVP, but it lacks GPU/large-scale training, advanced MLOps automation, and explicit profiling or quantization artifacts in public code. Some repository files include instructional prompts for writing style (AGENTS.md) and AI-authored markers; treat such files as metadata and not model-engineering guidance.
Phone
Technical skills
Languages
7
Python
C++
SQL
Java
C
MATLAB
Fortran
Python
3
Pydantic
FastAPI
Uvicorn
AI/ML
9
NumPy
Scikit-learn
Pandas
Statsmodels
Streamlit
SciPy
LLM
XGBoost
Jupyter Notebook
Analytics
3
Matplotlib
Plotly
Seaborn
Other
14
Azure
Hadoop
Spark
GitHub
Databricks
Git
Rest API
Docker
AI Agents
Machine Learning
CI/CD
RAG
Time Series
Anomaly Detection
Projects
Scientific Computing & Applied Research
Developed and modernized C/C++ and FORTRAN numerical software for nonlinear and dynamical systems, including high-order integration, parameter sweeps, stability analysis, and numerical verification. Used scientific computing techniques to validate simulation behavior and support applied mathematical research. Produced and contributed to peer-reviewed work in applied mathematics, computational modelling, imaging, and scientific computing.
Industrial AI
Developed predictive models and anomaly/risk monitoring for energy efficiency, abnormal heating, occupancy inference, and environmental risk using large multisensor datasets. Applied regression and inverse/diagnostic methods with calibration and uncertainty analysis to support operational decision-making. Added validation controls, decision thresholds, sensitivity tests, and safeguards to improve reliability, interpretability, and deployability.
WattVector / GridRisk AI
Built an end-to-end Python decision-support platform for wind/solar and net-demand forecasting with calibrated quantile predictions and scenario-based risk-aware decisions. Implemented deterministic and stochastic battery dispatch optimization using constraint handling and CVaR analysis. Packaged the solution into a Streamlit app with validation and technical-audit views and provided model interfaces via FastAPI with Docker-based reproducibility and automated testing.
Timeline
Senior Data Scientist / Research Lead (Applied Mathematics & AI)
•
Lead
Utopi Ltd
•
Full-Time
Lead end-to-end data science work from ambiguous business needs through data exploration, feature engineering, model development, validation, and production handover. Build predictive, forecasting, regression/classification, anomaly detection, and decision/risk models for large IoT time-series related to temperature, humidity, CO2, occupancy, and energy. Define modelling logic, evaluation criteria, safeguards, and validation controls, and communicate assumptions and uncertainty to stakeholders. Collaborate with engineering to operationalize models using Azure-oriented deployment patterns and production-grade APIs and testing.
Python
Azure
Glasgow Caledonian University
Master's Degree •
Big Data Technologies
Director of Research & Development
•
Head+
Smart and Scientific Solutions Ltd (SSS Techs)
•
Full-Time
Founded and led R&D across applied data science, mathematical modelling, scientific computing, and analytical software. Directed projects from quantitative problem formulation and prototype implementation to validation, documentation, and stakeholder presentations. Developed reproducible Python workflows for forecasting, optimization, simulation, risk analysis, and decision support.
Python
Senior AI/ML Engineer
Confidence: High Research
Quantitative finance researcher and applied statistics engineer (senior level) with demonstrated strengths in careful econometric inference and reproducible benchmarking for portfolio performance; the strongest proven skill is rigorous cross-sectional and time-series inference implemented in analysis/latent_performance.py and the HalfNormal SFA implementation in sfa/sfa_halfnormal.py. The codebase shows comprehensive unit and integration tests, deterministic report generation, and a deployable Streamlit/FastAPI MVP, but it lacks GPU/large-scale training, advanced MLOps automation, and explicit profiling or quantization artifacts in public code. Some repository files include instructional prompts for writing style (AGENTS.md) and AI-authored markers; treat such files as metadata and not model-engineering guidance.
Model Architecture & Training
6/10
How well models are designed and trained
Strong statistical model engineering and optimization work: custom likelihoods, analytic gradients, and careful numerical safeguards for non-trivial estimators.
Evidence
sfa/sfa_halfnormal.py: HalfNormalSFA._neg_loglik_grad, _loglike_obs, fit and analytic gradient-based L-BFGS optimisation
analysis/latent_performance.py: _profile_prior and hierarchical_shrinkage implementing (multivariate) empirical-Bayes profiling and shrinkage
Data Pipeline & Feature Engineering
5/10
How data is prepared for models
Robust data-loading and validation patterns with balanced-panel checks, design-matrix construction, and synthetic-data generators and validators.
Evidence
sfa/loaders.py: FactorDataset, design_matrix, load_portfolios and validate_factor_panel
moneybuffer/data_generation/households.py: generate_households and schema-aware synthetic data generation
moneybuffer/validation/input_validation.py: parse_household_csv and validate_household_inputs
Experimentation & Evaluation
5/10
How results are measured and tested
Well-structured evaluation and experiment code with forward validation, block bootstrap, HAC inference and extensive unit and integration tests enabling reproducible experiments.
Evidence
analysis/latent_performance.py: forward_performance_validation, summarize_forward_validation and block_bootstrap_rank_uncertainty
analysis/incremental_validation.py: run_incremental_test orchestration and _summarize_increment
tests/: multiple unit and integration tests (e.g., tests/test_quantitative_validity.py, tests/test_pipeline_integration.py) covering statistical properties and pipeline integrity
MLOps & Deployment
4/10
How models are shipped to production
Deployment-focused artifacts for a lightweight product: Streamlit front-end, FastAPI endpoint and PDF reporting; good test coverage for API endpoints and integration flows but limited production-grade MLOps automation shown.
Evidence
src/moneybuffer/api/main.py: FastAPI-based endpoints (HouseholdInput, ScoreResponse, scam_check)
app/streamlit_app.py and app/shared.py: Streamlit app wiring and session handling
reports/generate_report.py: deterministic PDF report generation using reportlab
Computational Efficiency
3/10
How efficiently computing resources are used
Some performance-aware choices (vectorised numpy, analytic gradients, eigenvalue flooring, finite-sample corrections) but no GPU/quantization/distributed-training or profiling artifacts.
Evidence
sfa/sfa_halfnormal.py: vectorised likelihood, analytic gradients, bounds and warm-start strategies
analysis/latent_performance.py: use of numpy linear algebra, eigen-decomposition and numerical eigenvalue floors in joint_newey_west_alpha_covariance
Research Depth & Innovation
6/10
Depth of research and new ideas
Solid research depth in applied econometrics and statistical inference: joint HAC covariance, multivariate empirical-Bayes, block bootstrap rank uncertainty and a novel incremental forward-test design.
Evidence
analysis/latent_performance.py: joint_newey_west_alpha_covariance, hierarchical_shrinkage, joint_alpha_tests
analysis/incremental_validation.py: the incremental rolling-vs-static forward test design and implementation
Expertise
MLOps & Model Lifecycle• Middle
Industries
Financial Services• Senior
Technologies
SQL
C++
MATLAB
Hadoop
Spark
Databricks
XGBoost
AI Agents
Azure• since 2025
CI/CD
Git
Docker• since 2026
LLM• since 2026
RAG• since 2026
Anomaly Detection• since 2026
Time Series
GitHub
Machine Learning
Recommendations
- Lead development of quantitative-statistics tooling and inference libraries for asset pricing and portfolio benchmarking (HAC, empirical-Bayes, block bootstrap).
- Build explainable fintech MVP features that combine rule-based scoring and lightweight ML (scam-detection pipelines, Streamlit front-end, FastAPI endpoints).
- Develop rigorous evaluation and reproducibility pipelines (deterministic bootstrap, forward validation, integration tests, deterministic report generation) for research-to-product workflows.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior Data Scientist
Confidence: Medium Statistician
A senior-level statistician and data-science practitioner focused on rigorous econometric inference and reproducible research pipelines; the most important strength is careful, uncertainty-aware statistical engineering. The strongest proven skill is applied statistical inference and reproducible pipeline construction, demonstrated by analysis/latent_performance.py (HAC covariance, empirical-Bayes, forward validation), analysis/incremental_validation.py (leakage-free incremental tests and block bootstrap), and sfa/sfa_halfnormal.py (robust SFA estimation). Public code does not primarily show large-scale distributed engineering (Spark/Hadoop) or production MLops at cloud scale, and there is limited evidence of high-throughput streaming system design or real-money trading infrastructure.
Statistical Rigor
8/10
Correct use of statistics
Strong statistical rigor: implements HAC covariance, joint Wald/GRS tests, Benjamini-Hochberg, empirical-Bayes profiling, stochastic-frontier inference, and explicit failure/assumption checks with unit tests and numerical safeguards.
Evidence
analysis/latent_performance.py: joint_newey_west_alpha_covariance, hierarchical_shrinkage, joint_alpha_tests, benjamini_hochberg
sfa/sfa_halfnormal.py: HalfNormalSFA class, likelihood, gradient, boundary test and AE computation
analysis/external_validation.py: validation checks comparing to statsmodels replication and reference manifests
Data Wrangling & Cleaning
8/10
Preparing and cleaning data
Good data wrangling and validation: dedicated loaders, balanced-panel checks, SHA-256 provenance, explicit normalization/parsing and clear error messages; tests exercise loader behaviors and failure modes.
Evidence
sfa/loaders.py: build_factor_dataset, design_matrix, validate_factor_panel, file_sha256
tests/test_loaders.py: tests for data loading, alignment and missing/required-columns behavior
Exploratory Analysis & Visualization
7/10
Exploring and visualizing data
Exploratory analysis and visualization are purposeful and interpretation-focused: figure-generation, table rendering and an integrated technical PDF report with narrative and derived-fact provenance rather than decorative plotting.
Evidence
analysis/figures.py: plotting functions and figure utilities (static_ae_ranking, ae_heatmap, etc.)
reports/generate_report.py: deterministic report assembly with captions, derived facts and figures
Predictive Modeling
8/10
Building models that predict
Predictive/modeling discipline is strong for both inferential and ML pieces: careful forward (leakage-free) validation, rolling/expanding-window tests, block bootstrap, HAC inference; a separate ML pipeline for scam detection with scikit-learn pipeline and training/evaluation utilities.
Evidence
analysis/incremental_validation.py: run_incremental_test, forward-per-window comparisons, _summarize_increment with HAC and block-bootstrap
analysis/latent_performance.py: forward_performance_validation and rolling_cross_sectional_performance
moneybuffer-uk/src/moneybuffer/scams/ml_model.py: scikit-learn pipeline, train_scam_classifier, predict_scam_ml
Business Insight & Impact
6/10
Turning analysis into business value
Good connection to business impact in two domains: careful framing of investment inference limits and a consumer-facing fintech MVP with explainable recommendations and action plans, though explicit cost-error tradeoff quantification is limited.
Evidence
reports/generate_report.py: explicit disclaimers, limitations, and practitioner-focused interpretation
moneybuffer-uk/src/moneybuffer/recommendations/action_engine.py: action plan generation and support link integration
Reproducibility & Notebook Hygiene
8/10
Clean, repeatable analysis
Reproducibility and hygiene are well addressed: pinned requirements, deterministic bootstrap seed, run manifest recording runtime and hashes, a top-level run_all orchestration, and a broad test suite exercising units and integration paths.
Evidence
reports/generate_report.py: references results/run_manifest.json and deterministic-seed usage
analysis/run_all.py: PipelineConfig, orchestration and end-to-end pipeline wiring; tests/test_pipeline_smoke.py
Expertise
Data Science• Senior
Industries
Financial Services• Senior
Technologies
Fortran
Jupyter Notebook
Scikit-learn
Seaborn
Matplotlib
SciPy
Plotly
Pandas
NumPy
Streamlit
Statsmodels
Recommendations
- Use to develop reproducible, uncertainty-aware quantitative research pipelines and academic-style technical reporting for asset-pricing or risk analysis.
- Implement forward-validation and bootstrap-based evaluation frameworks for forecasting systems where leakage and serial dependence matter.
- Build explainable consumer fintech features: scoring, rule+ML hybrid classifiers, and deterministic report/pdf generation with clear provenance and disclaimers.
- Lead development of rigorous statistical tooling (HAC inference, multivariate shrinkage, multiple-testing control) and tests for teams doing econometrics or causal inference.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior Backend Developer
Confidence: Medium Generalist
Senior-level Python backend generalist focusing on reproducible statistical analytics and consumer-facing prototypes; strongest at crafting numerically robust inference pipelines and clear module boundaries. The most proven skill is numeric statistics and reproducible analysis built in analysis/latent_performance.py and sfa/ (joint HAC covariance, empirical-Bayes shrinkage, block bootstrap) together with extensive tests. Public code shows little evidence of production-grade API hardening, authentication, secrets management or distributed systems instrumentation.
API Design
3/10
How well APIs are designed
Basic, correct API surface with Pydantic models and FastAPI endpoints and test coverage but missing higher-quality API design features such as versioning, idempotency handling, consistent error contract, auth, and pagination.
Evidence
moneybuffer-uk/src/moneybuffer/api/main.py: FastAPI app, Pydantic BaseModel schemas and endpoints (health, score_household, scam_check)
moneybuffer-uk/tests/test_api.py: integration tests for the API endpoints and validation behavior
Data Layer & Database
4/10
Working with databases
Thoughtful data-layer code with robust loaders, panel validation, and balanced-panel enforcement but no database migrations or explicit transaction/isolation logic; strong numeric correctness and data integrity checks instead of DB-level evolution artifacts.
Evidence
latent-performance-benchmarking/sfa/loaders.py: FactorDataset, dataset_manifest, file_sha256, validate_factor_panel and design_matrix utilities
latent-performance-benchmarking/analysis/latent_performance.py: pivoting residuals to balanced common-date panel and building joint HAC covariance
Scalability & Performance
3/10
Handling load and speed
Computationally efficient vectorised numeric implementations and a large bootstrap routine for statistical inference but limited evidence of system-level scalability patterns like caching strategies, queue-based decoupling, connection pooling, or measured load optimizations.
Evidence
latent-performance-benchmarking/analysis/latent_performance.py: block_bootstrap_rank_uncertainty implements a 5,000-draw circular block bootstrap with vectorised operations
latent-performance-benchmarking/sfa/sfa_halfnormal.py: numerically-stable likelihood, gradients and multiple warm-starts for optimizer performance
System Architecture
5/10
Overall system structure
Clear module separation and purposeful orchestration for analytics, modeling, reporting and a separate UI/API surface; good packaging and reproducibility design but not a distributed microservices decomposition with cross-service contracts.
Evidence
latent-performance-benchmarking/analysis/run_all.py: pipeline orchestration, CLI args, run manifest and wiring of analysis -> export -> report
moneybuffer-uk/src/moneybuffer: clear src layout separating api, resilience, scams, reporting and ui
Security & Auth
2/10
Protecting data and access
Some input validation and careful handling of data provenance are present, but minimal evidence of hardened authn/authz, token lifecycle, secrets handling or dependency/security-audit automation.
Evidence
moneybuffer-uk/src/moneybuffer/validation/input_validation.py: extensive input validation and CSV parsing logic
moneybuffer-uk/src/moneybuffer/api/main.py: public endpoints implemented but no auth or token handling
Reliability & Observability
5/10
Stability and monitoring
Strong test coverage, reproducibility artifacts, deterministic seeds and stability checks provide notable reliability and observability for numerical pipelines but limited runtime observability (structured logs, metrics, alerts) and operational resilience patterns (timeouts, retries, circuit-breakers) are visible.
Evidence
moneybuffer-uk/tests/: extensive unit and integration tests (including test_pipeline_integration.py)
latent-performance-benchmarking/reports/generate_report.py: writes and reads run_manifest.json; pipeline records runtime and hashes for reproducibility
Expertise
Python• Middle
Microservices & API Architecture• Middle
System Architecture• Middle
Industries
Data & Analytics• Senior
Financial Services• Middle
Health Care• Middle
Technologies
Python• since 2020 • Senior
Java
Rest API
FastAPI• since 2026
Pydantic
Uvicorn
Recommendations
- Develop statistically-intensive data services or analytics libraries that require rigorous numerical correctness and reproducibility (HAC inference, bootstrap, shrinkage).
- Package and productionise analytic pipelines into a data-service API with structured logging, metrics, and auth so models and reports can be served reliably under load.
- Build model-validation and CI pipelines that include dependency security checks, deterministic test harnesses, and reproducible run-manifests for regulated or high-assurance domains.
- Implement API versioning, idempotency, and explicit error contracts for consumer endpoints and add secrets/credential management before external deployment.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: