Inteshabul Haque
Data Analyst
Python
JavaScript
SQL
Data Pipeline & Feature Engineering: 5/10
Experimentation & Evaluation: 4/10
Active 3 days ago
+91 (8789) 796568
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
LLM-focused marketing analytics engineer at a mid-level who builds agent-backed analytics and CRO pipelines. The strongest proven skill is designing end-to-end marketing intelligence and analytics flows, evidenced by the Streamlit app (app/streamlit_app.py), A/B experiment analysis (backend/analytics/experiment_analysis.py) and the enterprise PySpark/XGBoost ML notebook. There is limited evidence of production-grade MLOps, automated model versioning, monitoring, or advanced deep learning research.
Technical skills
Python• Middle
JavaScript
SQL• Middle • 3y+
Python
FastAPI
pySpark
Databases
Delta Lake
FAISS
Databricks
AI/ML
Copilot
Scikit-learn
Spark
Pandas
NumPy
XGBoost
SHAP
Google GenAI SDK
Prophet• 3y+
Claude
dbt
Gemini
LangChain
Streamlit
Claude Code
Analytics
Plotly
Power BI• 3y+
Tableau
Frontend
React.js• 4y+
Timeline
Senior Associate, Intelligent Data Analytics & Cloud
•
Senior
PwC
•
Full-Time
Worked with Adobe’s Jarvis conversational AI team and Adobe Home on top-of-funnel and funnel analytics. Led conversation performance and intent capture analysis, including A/B testing automation to improve resolution and conversion. Built scalable analytics workflows using SQL, Databricks, Adobe Analytics, and Power BI for large-scale behavioral data. Developed dbt transformation models and AI-assisted reporting workflows using Claude, Streamlit, Gemini AI, and LangChain.
SQL
Databricks
Power BI
dbt
Claude
Streamlit
Gemini
LangChain
Assistant Manager, Business Analyst, Consumer Experience
•
Lead
Adidas
•
Full-Time
Analyzed large-scale user behavior data across 19 EU markets to support internal search, navigation, experimentation, and personalization. Drove CRO initiatives that contributed to significant revenue uplift and improved key engagement metrics, including reducing null search rates. Built reusable SQL data models and automated Power BI dashboards, using Prophet for forecasting and Amplitude for app behavior analysis. Collaborated with marketing and SEO teams to translate behavioral insights into campaign actions.
SQLsince 2023
Power BIsince 2023
Prophet
Amplitude
Assistant Manager, Digital Analytics & Site Operations
•
Lead
The Body Shop
•
Full-Time
Improved SEO performance through technical SEO analysis and crawl-focused investigations. Built live dashboards in Looker Studio and connected them with Search Console data for ongoing monitoring. Implemented and validated event tracking and the data layer using Google Tag Manager. Supported analytics implementation during a React.js migration.
React.js
E-commerce Analyst & Category Merchandiser
•
Middle
BulbulTV
•
Full-Time
Led seasonal campaign work and provided weekly behavioral insights to support merchandising decisions and category growth. Worked on conversion and engagement improvements, including enhancements aimed at increasing in-app CTR. Built dashboards to track performance and support ongoing optimization. Focused on translating user behavior signals into actionable merchandising and growth initiatives.
National Institute of Fashion Technology
Bachelor's Degree •
Apparel Production
Middle AI/ML Engineer
Confidence: Medium LLM Engineer
LLM-focused marketing analytics engineer at a mid-level who builds agent-backed analytics and CRO pipelines. The strongest proven skill is designing end-to-end marketing intelligence and analytics flows, evidenced by the Streamlit app (app/streamlit_app.py), A/B experiment analysis (backend/analytics/experiment_analysis.py) and the enterprise PySpark/XGBoost ML notebook. There is limited evidence of production-grade MLOps, automated model versioning, monitoring, or advanced deep learning research.
Model Architecture & Training
3/10
How well models are designed and trained
Practical ML training experience with classical models and statistical testing but no custom neural architectures or advanced training loops; shows correct use of XGBoost, train/test splits and hypothesis tests.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: XGBClassifier model.fit training blocks and predict/predict_proba usage
Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/experiment_analysis.py: calls to calculate_z_test and confidence_interval for A/B testing
Data Pipeline & Feature Engineering
5/10
How data is prepared for models
Clear data engineering and feature work including a medallion ETL pattern, PySpark transformations and explicit feature creation for ML models; also defensive dataset validation in analysis modules.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: Bronze -> Silver -> Gold tables and feature engineering (customer_gold, customer_ml_features)
Search_Heist_AI/analytics/dashboard_renderer.py and Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/funnel_analysis.py: dataset validation and funnel metric extraction
Experimentation & Evaluation
4/10
How results are measured and tested
Reasonable experimentation and evaluation practices: train/test splits, cross validation, grid search, metric reporting and SHAP explanations, but no experiment tracking, reproducible configs or automated A/B experiment pipelines.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: GridSearchCV, cross_val_score and printed evaluation metrics blocks
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: SHAP explainer and plots for model explainability
MLOps & Deployment
2/10
How models are shipped to production
Basic serving and app packaging present (FastAPI endpoints and Streamlit UI) but limited operational MLOps: no model versioning, no monitoring, simple file-based persistence and in-memory global state.
Evidence
Search_Heist_AI_CRO_Experimentation_Agent/backend/api/routes.py: FastAPI endpoints for upload and analysis and save_latest_dataset writing to datasets/latest_uploaded_file.csv
Search_Heist_AI/app/streamlit_app.py: Streamlit app integrating analysis and chat interface
Computational Efficiency
2/10
How efficiently computing resources are used
Some awareness of memory constraints and parallelism (data limiting, n_jobs=-1) and pragmatic sampling for explanations, but no GPU optimization, quantization, or advanced efficiency profiling.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: limits on data (df.limit(10000), sampling for SHAP) and GridSearchCV(n_jobs=-1)
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: memory-optimized recommendation variant with limited dataset
Research Depth & Innovation
1/10
Depth of research and new ideas
No original research contributions or novel model architectures; primarily practical application of well-known algorithms and libraries.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb: standard use of XGBoost, KMeans, Prophet and SHAP without custom model layers or research experiments
Expertise
AI / LLM Engineering (Agents)• Middle
RAG• Middle
Industries
Artificial Intelligence• Middle
Commerce• Middle
Sales & Marketing• Middle
Technologies
SQL• Middle • 3y+
Copilot
Claude
Spark
Databricks
dbt
Prophet• 3y+
Gemini
Anomaly Detection• mentioned only
Gemini• mentioned only
SQL• mentioned only
Recommendations
- Develop agent-driven analytics features and conversational marketing copilots that integrate dataset-aware prompts and funnel/experiment analysis.
- Build scalable ETL and feature pipelines for eCommerce analytics using PySpark and medallion patterns, and expand unit/integration tests around them.
- Implement basic MLOps practices: model serialization/versioning, reproducible training configs, and lightweight monitoring for deployed models and agents.
- Hardening production-grade LLM systems such as RAG with robust vector-store handling, retries/backoff, input sanitization and rate/cost controls.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: Medium Data Engineer
A mid-level data engineer and analytics developer focused on marketing and e-commerce analytics pipelines and dashboards. The strongest proven skill is building analytics pipelines and business-facing reporting as shown by modular analysis code such as analytics/experiment_analysis.py and analytics/funnel_analysis.py powering Streamlit and API endpoints. There is limited evidence of production-grade deployment, automated testing, or reproducible ML pipelines in public code.
Statistical Rigor
4/10
Correct use of statistics
Basic statistical tooling is implemented (z-tests, confidence intervals, uplift) and integrated into analysis flows, but there is limited evidence of formal assumption checks, multiple-comparison controls, or thorough uncertainty discussion beyond simple p-values and CIs.
Evidence
Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/statistics_engine.py:calculate_z_test, confidence_interval, calculate_uplift
Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/significance_test.py:run_significance_test
Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/experiment_analysis.py:analyze_experiment (uses z_test and p_value)
Data Wrangling & Cleaning
6/10
Preparing and cleaning data
Clear data ingestion and cleaning patterns are present with Spark medallion-style transforms in the notebook and defensive checks in analysis functions; some validation exists for required columns but broader data lineage, schema evolution, and automated data-quality pipelines are not evidenced.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb:Silver layer cleaning and dropDuplicates/fillna steps
Search_Heist_AI_CRO_Experimentation_Agent/backend/api/routes.py:save_latest_dataset (dataset persistence)
Search_Heist_AI_CRO_Experimentation_Agent/backend/analytics/funnel_analysis.py:analyze_funnel (required column validation)
Exploratory Analysis & Visualization
6/10
Exploring and visualizing data
Visualization and EDA are thoughtfully integrated into a Streamlit dashboard with programmatic chart generation and textual executive summaries that interpret metrics, although visual analysis often follows templated patterns rather than deep, hypothesis-driven exploration.
Evidence
Search_Heist_AI/analytics/chart_generator.py:generate_charts
Search_Heist_AI/analytics/executive_summary.py:generate_executive_summary
Search_Heist_AI/analytics/dashboard_renderer.py:render_dashboard (Streamlit integration)
Predictive Modeling
5/10
Building models that predict
Predictive modeling appears in a comprehensive notebook with XGBoost models, evaluation, SHAP explainability and hyperparameter search, indicating practical ML workflow skills; however, model pipelines are notebook-centric and lack productionization artifacts like training pipelines, versioning, or CI integration.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb:Churn prediction with XGBClassifier and evaluation
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb:SHAP explainability pipeline and hyperparameter GridSearchCV
Business Insight & Impact
6/10
Turning analysis into business value
Business-facing reasoning is a clear focus: executive summaries, KPI calculation, and conversational agent responses map technical outputs to business metrics and recommendations, showing good alignment to marketing/ecommerce use-cases, though direct cost/impact quantification per decision is limited.
Evidence
Search_Heist_AI/analytics/executive_summary.py:generate_executive_summary (business-focused insights)
Search_Heist_AI_CRO_Experimentation_Agent/backend/agents/master_agent.py:run_master_agent (business Q&A responses)
Reproducibility & Notebook Hygiene
3/10
Clean, repeatable analysis
Reproducibility is minimal: there is a requirements.txt and seeds in the notebook, but notebooks contain inline %pip, many manual steps, and there is no clear CI, DVC/data-versioning, or automated pipeline orchestration for reproducible runs.
Evidence
enterprise-ecommerce-analytics-ml-platform/enterprise-ecommerce-analytics-ml.ipynb:# NOTEBOOK META and inline %pip installs
Search_Heist_AI/requirements.txt and Search_Heist_AI_CRO_Experimentation_Agent/backend/requirements.txt (dependency lists)
Expertise
Analytics• Middle
Big Data• Middle
Industries
Commerce• Middle
Sales & Marketing• Middle
Technologies
Python• Middle
LangChain
Delta Lake
FAISS
Power BI• 3y+
SHAP
XGBoost
FastAPI
Scikit-learn
Google GenAI SDK
Plotly
Pandas
NumPy
pySpark
Streamlit
Analytics• mentioned only
Anomaly Detection• mentioned only
Gemini• mentioned only
SQL• mentioned only
Recommendations
- Develop production training and serving pipelines (CI/CD, model versioning, scheduled retraining) for the XGBoost models in the notebook
- Add automated data quality and schema checks (unit tests, Great Expectations) and data lineage for Spark/Delta workflows
- Extract core analysis logic from notebooks into reusable modules and add end-to-end integration tests and reproducible example scripts
- Introduce lightweight pipeline orchestration (Airflow/Prefect) and artifact/version tracking (MLflow/DVC) for deployment readiness
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Blockchain Developer
Confidence: Medium Generalist
A junior full-stack JavaScript developer focused on small web apps and AI integrations with a strength for building Express backends and DOM-driven frontends. The strongest proven skill is implementing an OpenAI-backed content pipeline and simple API surface, evidenced by server/services/openaiService.js and server/routes/archaeologyRoute.js. There is no public evidence of blockchain, smart contract, formal testing, or production-grade security and scalability practices.
Smart Contract Development
Writing smart contracts
Not evidenced in public code
Security & Audit
1/10
Checking contracts for flaws
Basic error handling and input validation exist but there is minimal security-specific engineering, no auditing artifacts, and potential risks when passing user text to an LLM without strict validation.
Evidence
ai-archaeologist-of-civilization/server/services/openaiService.js: try/catch around OpenAI call and fallback error response
ai-archaeologist-of-civilization/server/routes/archaeologyRoute.js: lightweight request validation (checks for topic)
ai-archaeologist-of-civilization/server.js: dotenv usage for credential management
Blockchain Protocol Understanding
Knowing how blockchains work
Not evidenced in public code
DApp & Web3 Integration
Connecting apps to blockchain
Not evidenced in public code
Tokenomics & DeFi Logic
Token and finance logic
Not evidenced in public code
Decentralization & Trust Model
Designing trust without middlemen
Not evidenced in public code
Industries
Artificial Intelligence• Middle
Commerce• Middle
Recommendations
- Build small full-stack prototypes that integrate third-party APIs and AI models (Express backend + React frontend).
- Harden the OpenAI integration by validating and sanitizing model outputs before JSON.parse and adding invariants and negative tests.
- Add unit and integration tests and a basic CI pipeline to increase reliability and team-readiness.
- Practise secure input handling and add rate limiting or request validation middleware for public endpoints.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: