Piyush Yadav
Analyst
SQL
Node JS
Python
Visual Basic
Data Pipeline & Feature Engineering: 4/10
Active 7 hours ago
+91 (8595) 955136
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
Backend-focused Node.js engineer at a middle level specializing in building REST APIs and JWT-based authentication. The strongest proven skill is practical authentication and refresh-token handling implemented in earnest/backend/src/services/auth.service.ts which includes bcrypt hashing, refresh-token storage and rotation. There is little public evidence of production-grade scaling, thorough automated tests, advanced observability, or multi-service distributed system design.
Technical skills
SQL
Node JS• Middle
Python• Middle
Visual Basic
Node JS
Bcrypt
Dotenv
Express
Prisma
Databases
MySQL
PostgreSQL
AI/ML
NumPy
Pandas
Prompt Engineering
Scikit-learn
XGBoost
TF-Keras
Deep Learning
SciPy
Jupyter Notebook
AI Agents
Edge AI
DevOps
Git
GitHub
Rest API
Analytics
Power BI
Tableau
Matplotlib
Seaborn
ETL/ELT
Timeline
Delhi Technological University (DTU)
Bachelor's Degree •
Software Engineering
Credit Risk Analysis
Backend
Analyzed credit-related data to identify risk patterns and factors affecting credit outcomes. Performed data cleaning, exploratory analysis, feature analysis, and statistical analysis to understand customer risk characteristics. Developed analytical insights to support risk assessment and data-driven decision-making.
Banking Fraud Detection
AI/ML
Analyzed banking transaction data to identify fraudulent patterns and understand key factors associated with fraudulent transactions. Performed data cleaning, preprocessing, exploratory data analysis, feature analysis, and machine learning to detect potential fraud. Evaluated model performance using relevant metrics and generated insights to support better fraud detection and risk monitoring.
Middle AI/ML Engineer
Confidence: High ML Engineer
ML engineer (mid-level) focused on applied supervised learning and time-series forecasting with practical strengths in data cleaning and end-to-end model prototyping. The strongest proven skill is data preparation and classical model evaluation, supported by concrete artifacts such as the credit risk notebook that contains intent-based imputation, age-based corrections, one-hot encoding and a model selection pipeline using XGBoost and GridSearchCV. There is limited evidence of production-grade MLOps, experiment tracking, unit tests or novel research contributions in public code.
Model Architecture & Training
3/10
How well models are designed and trained
Standard model building using classical ML and basic deep learning: implemented XGBoost/RF pipelines, GridSearchCV tuning and stacked LSTM models with Keras, but no custom losses, distributed training or advanced architecture innovations.
Evidence
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: classifiers list including XGBClassifier and model training loop with GridSearchCV
Piyushyadav10/Stock-Price-Prediction/LSTM_Improved_model(diff_dataset).ipynb: Sequential model with stacked LSTM and Dropout layers (model definition and model.fit)
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Concrete data work and feature engineering: careful cleaning, imputation strategies, one-hot encoding, date parsing and pivot aggregations used across datasets.
Evidence
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: intent-based 'Rate' imputation, age-based 'Emp_length' imputation, one-hot encoding of Home/Intent
Piyushyadav10/Marketing-Campaign-Analysis/Marketing Campaign Analysis.ipynb: date parsing, DOW feature, conversion_rate function and pivot tables for aggregation
Experimentation & Evaluation
3/10
How results are measured and tested
Basic experimentation and evaluation: train/val/test splits, common metrics (accuracy/precision/recall/AUC), model comparison and hyperparameter search are present, but no experiment tracking, reproducibility infrastructure or ablation studies.
Evidence
MLOps & Deployment
2/10
How models are shipped to production
Minimal deployment/MLOps artifacts: saved models are present, but there is no serving, versioning, monitoring, or CI/CD evidence.
Computational Efficiency
1/10
How efficiently computing resources are used
Little evidence of computational-efficiency engineering beyond using n_jobs and optional GPU flags; no quantization, profiling, mixed precision or batching optimizations.
Research Depth & Innovation
1/10
Depth of research and new ideas
Research and innovation depth is minimal; implementations follow standard patterns without novel algorithms or paper-reproduction rigor.
Expertise
Finance & FinTech AI• Middle
Marketing & AdTech• Middle
Industries
Financial Services• Middle
Sales & Marketing• Middle
Technologies
Deep Learning
SQL
MySQL
PostgreSQL
Prompt Engineering
Git
GitHub
Recommendations
- Develop credit-risk or scoring pipelines (feature engineering, model evaluation, and regulatory-aware reporting) using XGBoost/Tree models and clear validation sets.
- Build time-series forecasting components (data windows, scaling, Keras LSTM pipelines) for short-to-medium horizon predictions with careful backtesting.
- Lead analytical A/B testing and experimentation analyses, including automated segmentation, statistical tests and conversion-lift reports.
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 (approx middle level) focused on applied predictive analytics for credit risk, marketing measurement and time-series forecasting. The strongest proven skill is end-to-end applied modeling and analysis as shown by the credit risk pipeline with cleaning, targeted imputations, multiple classical models, grid search and final evaluation (Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb). What is not evidenced is production-grade engineering, MLOps, automated testing or robust reproducibility (no pinned environment, CI, or data/version control seen).
Statistical Rigor
3/10
Correct use of statistics
Basic statistical testing is present (t-tests for A/B), and standard metrics are used, but there is limited formal assumption checking, no multiple-comparison controls, and uncertainty calibration is minimal.
Data Wrangling & Cleaning
5/10
Preparing and cleaning data
Clear, hands-on data cleaning and imputation work with domain-aware choices (group-based imputations, outlier corrections) and dtype fixes; some heuristics are brittle but demonstrate practical data-wrangling skill.
Evidence
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: imputation of Rate by loan Intent and Emp_length imputed by Age; one-hot encoding of categorical fields
Piyushyadav10/Marketing-Campaign-Analysis/Marketing Campaign Analysis.ipynb: date parsing, type fixes, creation of DOW_subscribed and language-correctness columns
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Extensive EDA and visualization with written interpretation after many plots; pairplots, heatmaps and per-segment charts are used and observations are recorded.
Evidence
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: histograms, boxplots, pairplots, heatmap plus narrative observations
Piyushyadav10/Marketing-Campaign-Analysis/Marketing Campaign Analysis.ipynb: time-series plots, conversion-rate bar charts, day-of-week analyses with interpretive notes
Predictive Modeling
4/10
Building models that predict
Multiple supervised models and tuning workflows are implemented (grid search, PyCaret compare/tune), plus deep-learning LSTM prototypes; however cross-validation and tuning use standard recipes without nested CV, limited error analysis and no clear productionization steps.
Evidence
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: training of LogisticRegression, RandomForest, XGBoost plus GridSearchCV and final evaluation
Piyushyadav10/Credit-Risk-Analysis/PyCaret/credit_risk_analysis_PyCaret.ipynb: compare_models, tune_model and finalize_model workflows
Piyushyadav10/Stock-Price-Prediction/LSTM_Improved_model(diff_dataset).ipynb: multi-layer LSTM model, training loop, model.save('keras_model.h5')
Business Insight & Impact
4/10
Turning analysis into business value
Analyses often connect to business questions and produce actionable findings (e.g., identifying a language bug and estimating lost conversions), but cost-of-error tradeoffs and operational impact quantification are limited.
Evidence
Piyushyadav10/Marketing-Campaign-Analysis/Marketing Campaign Analysis.ipynb: investigation of House Ads language bug and expected lost conversions calculation
Piyushyadav10/Credit-Risk-Analysis/credit_risk_analysis_Scikit.ipynb: reporting of precision/recall/specificity and discussion of model performance for lending decisions
Reproducibility & Notebook Hygiene
2/10
Clean, repeatable analysis
Some reproducibility steps exist (saving models, PyCaret session_id), but there is no pinned environment, requirements, CI, data versioning, or systematic notebook-to-module hygiene.
Industries
Financial Services• Middle
Sales & Marketing• Middle
Technologies
Python• Middle
Jupyter Notebook
XGBoost
Scikit-learn
Seaborn
Matplotlib
SciPy
Pandas
NumPy
TF-Keras
Recommendations
- Develop production-ready model pipelines: add environment files, requirements, lightweight unit tests, data versioning and modularize notebooks into reusable scripts.
- Work on robust validation: adopt nested CV or holdout strategies for hyperparameter tuning, explicitly check assumptions for statistical tests, and produce calibration and uncertainty estimates.
- Lead end-to-end analytics prototypes: A/B test analysis automation, instrumentation to quantify business impact and cost-based error analysis for credit decisions.
- Harden data cleaning: replace brittle manual fixes with rule-based validation, anomaly detection and logging so domain experts can sign off on heuristic corrections.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Backend Developer
Confidence: High API Engineer
Backend-focused Node.js engineer at a middle level specializing in building REST APIs and JWT-based authentication. The strongest proven skill is practical authentication and refresh-token handling implemented in earnest/backend/src/services/auth.service.ts which includes bcrypt hashing, refresh-token storage and rotation. There is little public evidence of production-grade scaling, thorough automated tests, advanced observability, or multi-service distributed system design.
API Design
5/10
How well APIs are designed
API design shows deliberate request/response patterns, consistent success/error utilities, pagination and filtering, and token endpoints with refresh/logout, but lacks explicit versioning, idempotency keys and rate limiting.
Evidence
earnest/backend/src/controllers/auth.controller.ts
earnest/backend/src/controllers/task.controller.ts
earnest/backend/src/services/task.service.ts
Data Layer & Database
4/10
Working with databases
Uses Prisma ORM with guarded DB operations and Prisma-specific error handling; demonstrates parallelized reads (count + findMany) and refresh-token storage, but no migration_chain or explicit transaction/isolation handling is shown.
Evidence
earnest/backend/src/services/auth.service.ts
earnest/backend/src/services/task.service.ts
earnest/backend/src/middleware/error.ts
Scalability & Performance
3/10
Handling load and speed
Some performance awareness (parallelizing count and findMany) and a front-end perf hook exist, but there is no caching strategy, queueing, connection-pooling tuning, or load-test/benchmark artifacts.
Evidence
earnest/backend/src/services/task.service.ts
nested-question-form/src/reportWebVitals.js
System Architecture
4/10
Overall system structure
Code is modular with clear separation (controllers, services, middleware, prisma client), showing conventional monolith service decomposition, but there is no evidence of multi-service contracts, service discovery, or advanced config/secret orchestration.
Evidence
earnest/backend/src/controllers/task.controller.ts
earnest/backend/src/services/auth.service.ts
earnest/backend/src/middleware/error.ts
Security & Auth
5/10
Protecting data and access
Solid practical auth choices are implemented (bcrypt with explicit salt rounds, JWT with refresh-token rotation stored in DB, logout/revocation), and JWT errors are handled, but input validation and broader dependency/audit practices are not fully visible.
Evidence
earnest/backend/src/services/auth.service.ts
earnest/backend/src/middleware/error.ts
earnest/backend/src/middleware/auth.ts
Reliability & Observability
3/10
Stability and monitoring
Global error handling, AppError class and controller try/catch usage show basic reliability and observability thought, but there is no structured logging, metrics, retry/backoff strategies or graceful shutdown logic visible.
Evidence
earnest/backend/src/middleware/error.ts
earnest/backend/src/controllers/task.controller.ts
Expertise
Node.js• Middle
Microservices & API Architecture• Middle
Databases & Vector Storage• Middle
Technologies
Node JS• Middle
Rest API
Express
Bcrypt
Prisma
Dotenv
Recommendations
- Develop and harden backend REST endpoints (auth, tasks) into production-ready services with input validation, request schemas and automated tests.
- Add schema migration history and documented migration_chain steps along with CI hooks to run migrations and prisma generate during deploys.
- Introduce observability and reliability primitives: structured logs, request correlation ids, metrics (Prometheus) and graceful shutdown handling in server.ts.
- Implement rate limiting, caching strategy (with invalidation), and a basic load-testing benchmark to guide performance improvements.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: