Bhanwar Singh Rathore
Machine Learning Engineer
Python
SQL
Data Pipeline & Feature Engineering: 4/10
Active 1 day ago
+91 (8946) 872618 Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
Data engineer in the middle experience band specializing in building retrieval-augmented and agentic LLM prototypes and document ingestion pipelines. The strongest proven skill is assembling RAG retrieval and embedding pipelines as shown by the Chroma embedding ingestion and retriever usage in data_injection.py and the rag_generation.py chain. The work lacks evidence of production hardening such as automated tests, monitoring, input validation for PII-sensitive medical data, and rigorous evaluation methodology.
Technical skills
Python• Middle
SQL• Junior
AI/ML
BERT
Computer Vision
Embeddings
Fine-tuning
Gemini
Google Colab
Grok
Groq
LangChain
LangGraph
LlamaIndex
LLM
LoRA
Multimodal AI
NumPy
Ollama
Pandas
PEFT
Prompt Engineering
QLoRA
RAG
Semantic Search
Streamlit
Tokenization
Google GenAI SDK
huggingface_hub
Scikit-learn
NLP
PyTorch
TensorFlow
Transformers
DevOps
Vector
Analytics
Power BI
Tableau
Matplotlib
Seaborn
Timeline
University of Cambridge
Master's Degree •
Data Science and Gen AI
Data Science Intern
•
Junior
Next Hikes IT Solutions
•
Internship
Developed and evaluated machine learning models using Python and scikit-learn, supporting predictive analytics workflows. Worked with datasets of 10,000+ records, performing data processing and analysis with Python and SQL. Implemented feature engineering, model optimization, and evaluation, and collaborated on AI and analytics projects.
Python
Scikit-learn
SQL
AI/ML Projects
AI/ML
RAG
Python
LLM
Keras
Chroma
SQL
Scikit-learn
Seaborn
Speech Recognition
Streamlit
huggingface_hub
Middle AI/ML Engineer
Confidence: Medium LLM Engineer
LLM-focused engineer (mid-level) who builds practical RAG and LLM-powered prototypes with simple UI frontends; strongest at integrating hosted LLMs and vector search into working pipelines. The clearest proven skill is assembling a working RAG pipeline and Streamlit interfaces using Chroma persistence and HuggingFace embeddings, exemplified by data_injection.py and rag_generation.py. There is little evidence of custom model training, experiment tracking, performance engineering, or production MLOps in public code.
Model Architecture & Training
1/10
How well models are designed and trained
No custom model architecture or training loops; only API usage of hosted LLMs and setting of basic LLM parameters.
Evidence
Project-10/Project_10_medical_report_extraction.ipynb: genai.GenerativeModel("gemini-2.5-flash") usage
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/rag_generation.py: ChatGroq(...) llm instantiation with max_tokens=512
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Reasonable data ingestion and preprocessing for RAG and OCR projects, including PDF loading, text splitting and basic dataset handling for images.
Evidence
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/data_injection.py: PyPDFDirectoryLoader and RecursiveCharacterTextSplitter usage with chunk_size/chunk_overlap
Project-10/Project_10_medical_report_extraction.ipynb: image file discovery, deterministic random sampling into training/testing splits
Experimentation & Evaluation
1/10
How results are measured and tested
Minimal experimentation and evaluation artifacts; only simple prints and a small train/test split in a notebook without metrics, tracking or reproducible experiment harness.
Evidence
Project-10/Project_10_medical_report_extraction.ipynb: random.sample selection for training/testing and printing of results
MLOps & Deployment
2/10
How models are shipped to production
Basic deployment-oriented pieces: Streamlit frontends and persistent vector DB usage, but no production serving, versioning or monitoring apparatus.
Evidence
Project-9/app.py and RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/app.py: Streamlit apps for UX/serving prototypes
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/data_injection.py: Chroma.from_documents with persist_directory
Computational Efficiency
1/10
How efficiently computing resources are used
Little to no effort in computational efficiency beyond basic chunking and conservative max_tokens; no batching, quantization, or GPU profiling evidence.
Evidence
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/data_injection.py: text chunking parameters (chunk_size/chunk_overlap)
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/rag_generation.py: llm configured with max_tokens=512
Research Depth & Innovation
1/10
Depth of research and new ideas
No research-level artifacts or novel algorithmic work; uses established libraries and prompt templates without ablations or reproduced results.
Evidence
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/rag_generation.py: use of ChatPromptTemplate and straightforward RAG chain construction
Expertise
RAG• Middle
AI / LLM Engineering (Agents)• Middle
Industries
Health Care• Middle
Technologies
Python• Middle
SQL• Junior
LangGraph
Grok
Groq
LlamaIndex
LoRA
Fine-tuning
Embeddings
Scikit-learn
Prompt Engineering
Multimodal AI
Computer Vision
NLP
Google GenAI SDK
Ollama
PEFT
QLoRA
Google Colab
Transformers
TensorFlow
NumPy
PyTorch
Gemini
BERT
Vector
Tokenization
Semantic Search
AWS• mentioned only
Cloud• mentioned only
Recommendations
- Develop RAG-based document QA and retrieval prototypes, extending current pipeline with tests, metrics and automated data ingestion.
- Build polished Streamlit/UX prototypes for LLM features and harden them with input validation, rate limiting and simple guardrails.
- Migrate prototype components into reproducible MLOps flows (CI/CD for vector DB builds, model-call wrappers with retries/backoff and logging).
- Add evaluation and experiment tracking (W&B or MLflow), and introduce basic performance work such as batching or quantization when moving to production.
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
Data engineer in the middle experience band specializing in building retrieval-augmented and agentic LLM prototypes and document ingestion pipelines. The strongest proven skill is assembling RAG retrieval and embedding pipelines as shown by the Chroma embedding ingestion and retriever usage in data_injection.py and the rag_generation.py chain. The work lacks evidence of production hardening such as automated tests, monitoring, input validation for PII-sensitive medical data, and rigorous evaluation methodology.
Statistical Rigor
Correct use of statistics
Not evidenced in public code
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Contains end-to-end document ingestion and text chunking into a vector DB, basic image processing and JSON extraction, and simple error handling for external queries.
Evidence
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/data_injection.py: PyPDFDirectoryLoader, RecursiveCharacterTextSplitter, Chroma.from_documents pipeline
Project-10/Project_10_medical_report_extraction.ipynb: image loading, prompt-based extraction, JSON parsing and CSV/Excel export
Exploratory Analysis & Visualization
2/10
Exploring and visualizing data
Minimal exploratory analysis or visualization; notebook shows basic DataFrame creation and head but no written interpretation or multi-step EDA.
Evidence
Project-10/Project_10_medical_report_extraction.ipynb: pd.DataFrame(records), df.head()
Predictive Modeling
2/10
Building models that predict
Uses pretrained LLMs and embedding models and builds a RAG retrieval pipeline, but there is no model training, evaluation methodology, or proper error analysis.
Evidence
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/rag_generation.py: HuggingFaceEmbeddings, Chroma retriever and ChatGroq LLM chain
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/data_injection.py: embedding generation and vector store persistence
Business Insight & Impact
2/10
Turning analysis into business value
Some structure for producing business-facing outputs (report templates and PDF export) but no explicit business metric framing, cost-of-error reasoning, or product-level impact analysis.
Evidence
Pharma-Multi-Agent-System/report_agent.py: formatted final report prompt and structure
Pharma-Multi-Agent-System/pdf_export.py: export_pdf producing downloadable PDF reports
Reproducibility & Notebook Hygiene
2/10
Clean, repeatable analysis
Basic reproducibility signals such as fixed random seed and Chroma persistence exist, but there is no pinned environment, CI, tests, data versioning or robust config management.
Evidence
Project-10/Project_10_medical_report_extraction.ipynb: random.seed(842) and explicit pip installs
RAG-Based-Document-Question-Answering-System-LangChain-ChromaDB-Groq-LLM/rag_generation.py: load_dotenv and persist_directory usage for Chroma
Expertise
Analytics• Middle
Industries
Health Care• Middle
Technologies
Tableau
LangChain
Power BI
Seaborn
Matplotlib
huggingface_hub
Pandas
LLM
RAG
Streamlit
Recommendations
- Develop production hardening: add unit/integration tests, input validation and PII-handling logic for medical data, and CI to ensure safe deployments.
- Add evaluation and error analysis pipelines: logging of retrieval hits, human-review annotations, and automated metrics to measure RAG accuracy and hallucination rates.
- Introduce environment and data reproducibility: pin dependencies (requirements or poetry), add DVC or dataset versioning for medical images and vector DB snapshotting.
- Harden security and secrets management: remove in-notebook API key access patterns and use secure secret stores and runtime-config patterns.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: