Shivam Singh
Software Developer
Python
Java
SQL
JavaScript
Model Architecture & Training: 4/10
Data Pipeline & Feature Engineering: 4/10
Active 11 days ago
Invite to interview
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A backend API engineer working in Java and Spring with a focus on e-commerce inventory and order-pricing logic at a mid-senior level, with the strongest practical skill being modular domain design for pricing and packaging. The most proven technical asset is the rule-based PricingEngine and ordered PricingRule implementations together with automated tests such as PricingEngineTest and StoreOrderIntegrationTest. There is limited evidence of production-grade observability, explicit DB migration history, and advanced resilience patterns such as retries with jitter or circuit breakers in public code.
Phone
Technical skills
Languages
4
Python
Java
SQL
JavaScript
Java
7
Spring Boot
Spring Data JPA
Lombok
Spring Security
Jackson
Spring MVC
AssertJ
DevOps
6
GitHub
AWS
Amazon S3
Kubernetes
Amazon EC2
AWS Lambda
Databases
4
OpenSearch
PostgreSQL
MySQL
ElasticSearch
AI/ML
9
NumPy
Pandas
Scikit-learn
OpenCV
XGBoost
Keras
TF-Keras
LLM
Claude
Other
15
Matplotlib
React.js
Seaborn
Docker
Rest API
Copilot
CI/CD
AI Agents
Deep Learning
AI/ML
GraphQL
RAG
Computer Vision
Prompt Engineering
Image Segmentation
Timeline
Software Engineer I
•
Middle
Planview
•
Full-Time
Worked on backend services supporting enterprise workflow functionality, focusing on improving approval processing performance. Implemented optimizations such as query reduction and hash-based caching to speed up critical paths. Owned automation and troubleshooting of data archival and search reliability components, including resolving OpenSearch issues. Participated in technical code reviews using AI-assisted tooling.
OpenSearch
Lovely Professional University (LPU)
Bachelor's Degree •
Computer Engineering
Python Developer Intern
•
Junior
Trailytics
•
Internship
Built automated data extraction pipelines to process large volumes of scraped data for downstream use. Improved data consistency and integrity across scraping and integration pipelines. Delivered pipeline outputs suitable for further analytics by cleaning and structuring incoming data.
Python
Middle AI/ML Engineer
Confidence: Medium ML Engineer
A mid-level ML engineer focusing on applied computer vision and tabular ML with a strongest practical skill in building UNet-based medical image segmentation pipelines. The clearest proven artifact is the UNet training and inference pipeline - including 3D NIfTI slicing, multiprocessing preprocessing and dice-loss training in the MRI notebooks. Not evidenced are rigorous experiment tracking, production-grade serving and formal reproducibility or test suites across projects.
Model Architecture & Training
4/10
How well models are designed and trained
Practical model architecture and training code - custom U-Net, dice loss and end-to-end training - but limited advanced regularization, ablation or rigorous training engineering.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: def unet
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: def dice_loss
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: model.fit_generator and model.save('UNET-BRAINCTA_... .h5')
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Concrete 3D medical-image preprocessing and simple tabular feature engineering; uses multiprocessing for slicing and Keras data generators for training.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: sliceAndSaveVolumeImage and sliceAndSaveSlice using multiprocessing.Pool
PROXMED-LPU-Hackathon-August-2023/Image_Segmentation.ipynb: create_segmentation_generator_train using ImageDataGenerator
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: df['Total_Bill'] feature creation and preprocessing steps
Experimentation & Evaluation
2/10
How results are measured and tested
Basic training/validation flows and metric reporting are present but there is no structured experiment tracking, reproducible config or formal evaluation pipeline.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: model.compile(metrics=[MeanIoU]) and model.fit_generator logs
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: imports of GridSearchCV and cross_val_score indicating some model selection intent
MLOps & Deployment
3/10
How models are shipped to production
Minimal deployment artefacts - model save/load and a Streamlit app for inference - but no production serving, CI/CD or model lifecycle/versioning evidence.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: model.save('UNET-BRAINCTA_{IMAGE_HEIGHT}_{IMAGE_WIDTH}.h5')
Customer-Churn-Prediction/app.py: imports streamlit and pickle indicating a lightweight inference app
Computational Efficiency
2/10
How efficiently computing resources are used
Some attention to throughput in preprocessing (parallel slicing) and batching, but no profiling, quantization or GPU/memory optimization evidence.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: use of multiprocessing.Pool(cpu_count()) to parallelize slice saving
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb and Image_Segmentation.ipynb: explicit BATCH_SIZE constants and manual IMG resizing in predictVolume
Research Depth & Innovation
1/10
Depth of research and new ideas
No evidence of novel research, new algorithmic contributions or reproduced SOTA results; implementations follow standard, well-known patterns.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: unet implementation matches standard U-Net pattern
Image_Segmentation.ipynb: textbook dice_loss and standard training loop
Verified artifacts
Expertise
Computer Vision & Image Analysis• Middle
Medical AI & Healthcare• Middle
Industries
Health Care• Middle
Telecommunications• Middle
Technologies
SQL
MySQL
PostgreSQL
Copilot
Claude
Prompt Engineering
Computer Vision
AI Agents
CI/CD
Pandas
Keras
AWS
Docker
Kubernetes
ElasticSearch
LLM
RAG
OpenSearch• since 2023
AWS Lambda
Amazon EC2
GitHub• since 2008
Amazon S3
Recommendations
- Develop medical-image segmentation prototypes and PoCs that convert NIfTI volumes to robust 2D/3D training pipelines and evaluation scripts.
- Build end-to-end model training and evaluation pipelines with experiment tracking (W&B or MLflow), config-driven runs and reproducible notebooks to improve reproducibility.
- Implement lightweight deployment for inference with containerized serving (FastAPI/TorchServe/Triton) and add CI/CD for model updates rather than only Streamlit frontends.
- Add automated data validation and unit tests around preprocessing and a clear evaluation / holdout protocol for clinical metrics and tabular models.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: Medium ML Practitioner
An ML practitioner at a middle level specializing in applied deep learning and prototyping predictive models for both medical image segmentation and tabular churn prediction. The strongest proven skill is building and training UNet-based segmentation pipelines and volume-level prediction logic as implemented in PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb with a custom dice loss and predictVolume workflow. There is little evidence of production hardening, automated testing, data versioning or rigorous statistical validation in public artifacts.
Statistical Rigor
2/10
Correct use of statistics
Basic descriptive statistics and model metrics are present but there is little evidence of formal statistical testing, uncertainty quantification, multiple-comparison controls or rigorous evaluation protocols.
Evidence
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: descriptive statistics and correlation heatmap
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: training logs showing Dice Loss and MeanIoU without cross-validation or uncertainty analysis
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Concrete data wrangling for domain-specific formats is implemented, including NIfTI reading, normalization and multi-axis slicing, and tabular cleaning steps are visible, but there is limited evidence of robust pipeline protections like provenance, data versioning or leakage checks.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: readImageVolume, normalizeImageIntensityRange, sliceAndSaveVolumeImage, imageVolumes
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: df.drop, creation of Total_Bill, null checks
Exploratory Analysis & Visualization
4/10
Exploring and visualizing data
Exploratory visualizations and domain-focused displays exist and are used to inspect results, and image display utilities are implemented, but written interpretation is limited and charts are mostly exploratory rather than structured storytelling tied to decisions.
Evidence
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: pie chart, heatmap, boxplots and inline EDA cells
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: display and show_prediction functions with image/mask visualizations
Predictive Modeling
4/10
Building models that predict
Solid applied modeling work is present: a custom UNET implementation, custom dice loss, model training and volume-level prediction logic; churn notebook shows multiple classical models and hyperparameter tuning imports. Missing are rigorous validation strategies, explicit baseline-first discipline, calibration analysis and systematic error analysis.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: unet(n_levels, ...) function, dice_loss, model.compile and model.fit_generator, predictVolume
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: imports for LogisticRegression, RandomForest, XGBClassifier, GridSearchCV and kerastuner RandomSearch
Business Insight & Impact
2/10
Turning analysis into business value
There is a stated business objective for churn prediction but little evidence of concrete business-metric translation, cost-of-error analysis, or actionable recommendations derived from model outputs.
Evidence
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: objective description of churn prediction
Customer-Churn-Prediction/README.md: description of models tried and metrics used without FP/FN cost analysis
Reproducibility & Notebook Hygiene
3/10
Clean, repeatable analysis
Notebooks save models and define a SEED constant and there are reproducible training calls, but reproducibility practices are weak: hardcoded Google Drive paths, ad-hoc pip installs in cells, garbled requirements file and no CI, tests or data/version management.
Evidence
PROXMED-LPU-Hackathon-August-2023/MRI (1).ipynb: SEED constant, model.save and use of custom_object_scope
Customer-Churn-Prediction/Customer_Churn_Prediction.ipynb: inline pip installs, absolute paths to data files
Industries
Health Care• Middle
Telecommunications• Middle
Technologies
AI/ML
Deep Learning
Python• since 2023 • Middle
OpenCV
XGBoost
Scikit-learn
Seaborn
Matplotlib
NumPy
TF-Keras
Image Segmentation
Recommendations
- Develop production-ready segmentation pipelines by modularizing the notebook code into reusable Python modules, adding unit tests and containerized reproducible runs (Docker) for training and inference.
- Harden model evaluation: add cross-validation, uncertainty quantification, calibration checks, and structured error analysis for both segmentation and churn models.
- Convert the churn notebook prototypes into a repeatable ML pipeline with clear feature lineage, leakage checks and a baseline-first benchmark framework.
- Improve reproducibility and deployment readiness by creating a clean, pinned requirements file, removing hardcoded cloud drive paths, and adding simple CI or run scripts to reproduce training end-to-end.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Backend Developer
Confidence: Medium API Engineer
A backend API engineer working in Java and Spring with a focus on e-commerce inventory and order-pricing logic at a mid-senior level, with the strongest practical skill being modular domain design for pricing and packaging. The most proven technical asset is the rule-based PricingEngine and ordered PricingRule implementations together with automated tests such as PricingEngineTest and StoreOrderIntegrationTest. There is limited evidence of production-grade observability, explicit DB migration history, and advanced resilience patterns such as retries with jitter or circuit breakers in public code.
API Design
4/10
How well APIs are designed
Reasonable REST API design with DTO validation and a global exception handler, but no visible versioning, idempotency keys, or pagination strategy.
Evidence
2000shivam659/duck-store/demo/src/main/java/com/duckstore/store/controller/StoreOrderController.java
2000shivam659/duck-store/demo/src/main/java/com/duckstore/warehouse/controller/DuckWarehouseController.java
2000shivam659/duck-store/demo/src/main/java/com/duckstore/common/exception/GlobalExceptionHandler.java
Data Layer & Database
5/10
Working with databases
Use of JPA entities and repositories with explicit service transaction boundaries and isolation awareness; no migration history present though.
Scalability & Performance
4/10
Handling load and speed
Some concurrency testing and modular pricing components suggest attention to scalability, but there is no evidence of caching strategies, queue-based decoupling, or measured performance tuning.
System Architecture
5/10
Overall system structure
Clear package separation and a modular rule-based pricing engine show deliberate architectural choices for extensibility, though the project is a monolith and not a multi-service system.
Security & Auth
2/10
Protecting data and access
Basic input validation and HTTP error mapping exist, but there is little evidence of authentication/authorization, token lifecycle management, secrets hygiene, or dependency-audit integration.
Reliability & Observability
4/10
Stability and monitoring
Good unit and integration test coverage including concurrency tests and integration tests, but limited evidence of observability, structured correlation ids, retries with backoff, or circuit breakers.
Expertise
Java• Middle
Microservices & API Architecture• Middle
Industries
Commerce• Middle
Technologies
Java• since 2023 • Middle
Rest API
Jackson
GraphQL
Spring Boot
Spring MVC
Spring Data JPA
Spring Security
Lombok
AssertJ
Recommendations
- Implement database migration tooling (Flyway or Liquibase) and publish migration history to demonstrate safe schema evolution.
- Add structured logging with request correlation ids and integrate basic metrics/alerts to improve observability and incident response.
- Hardening resilience at service boundaries by adding timeouts, retries with exponential backoff and circuit-breaker patterns where necessary.
- Extract or document operational deployment artifacts (health checks, readiness/liveness probes, connection pool tuning) to prepare the app for production traffic.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: