Anshu Vaish
Data Analyst
Python
SQL
Data Wrangling & Cleaning: 4/10
Active 10 hours ago
+91 (7007) 242195
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A practical ETL-focused data engineer (middle level) who builds small end-to-end pipelines that enrich RSS news with LLM extractions and deliver deduplicated results to stakeholder spreadsheets. The strongest proven skill is modular ETL and light productionization shown by the extract/transform/load scripts and the load.py deduplication and Google Sheets push. The codebase lacks automated testing, robust error handling, monitoring, and evidence of large-scale or high-throughput data engineering work.
Technical skills
Python• Senior • 3y+
SQL• Senior • 3y+
Python
Beautiful Soup
Databases
Google BigQuery
LookML
PostgreSQL
Snowflake
AI/ML
Google GenAI SDK
LLM Apps
NumPy• 3y+
Pandas• 3y+
LLM
Claude
Gemini
DevOps
CI/CD
GCP
Git
SLI/SLO/SLA
Rest API
Analytics
Matplotlib• 3y+
Seaborn• 3y+
Power BI
QA
Selenium
Timeline
Data Analyst
•
Middle
H.G. Technologies
•
Full-Time
Designed normalized data models and entity structures to support a single source of truth across reporting environments. Built interactive Power BI dashboards using complex measures and optimized SQL for operational KPI and SLA reporting. Automated ETL-related workflows and incident handling by integrating Power Automate/Power Query with REST APIs and Microsoft Teams/Outlook. Implemented Python-based extraction pipelines and an LLM-assisted extraction and validation workflow for unstructured documents.
Power BI
SQL
Rest API
Python
Beautiful Soup
Selenium
Pandas
LLM
JavaScript
Data Analyst
•
Middle
Be Me
•
Full-Time
Delivered real-time executive BI dashboards in Looker Studio to track sales, operations, and customer metrics. Improved supply-chain analytics and ETL/ELT performance by applying Python data processing and SQL query optimization. Built an automated invoice workflow and improved data quality by reducing manual reporting errors. Enhanced retention insights using RFM segmentation and governed data structures for loyalty targeting.
SQLsince 2023
Pythonsince 2023
NumPy
Pandassince 2023
LookML
Associate Data Analyst
•
Middle
Epic Businesses
•
Full-Time
Performed sales territory identification using dimensional modeling and geographic segmentation to support BI strategy decisions. Streamlined reporting by creating automated dashboards with macros and exploratory analysis workflows. Used visualization in Python to reduce multi-hour reporting efforts into one-click outputs.
Matplotlib
Allahabad University
Bachelor's Degree •
Mathematics
Middle Data Scientist
Confidence: Medium Data Engineer
A practical ETL-focused data engineer (middle level) who builds small end-to-end pipelines that enrich RSS news with LLM extractions and deliver deduplicated results to stakeholder spreadsheets. The strongest proven skill is modular ETL and light productionization shown by the extract/transform/load scripts and the load.py deduplication and Google Sheets push. The codebase lacks automated testing, robust error handling, monitoring, and evidence of large-scale or high-throughput data engineering work.
Statistical Rigor
1/10
Correct use of statistics
Almost no statistical or inferential work; the pipeline extracts sentiment and category labels via an LLM but contains no assumption checks, hypothesis tests, uncertainty quantification or negative-result discussion.
Evidence
ai-market-intelligence-pipeline/transform.py: prompt and JSON schema request a 'sentiment' field from the LLM
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Basic data wrangling is present: parsing RSS, constructing DataFrames, date normalization and simple deduplication, but there is limited provenance, validation, robust missing-value strategies or leakage prevention beyond simple fills and drop_duplicates.
Evidence
ai-market-intelligence-pipeline/extract.py: fetch_unstructured_data builds DataFrame from RSS entries
ai-market-intelligence-pipeline/transform.py: published_date normalization with pd.to_datetime
ai-market-intelligence-pipeline/load.py: fillna and drop_duplicates(subset=['link']) before upload
Exploratory Analysis & Visualization
1/10
Exploring and visualizing data
Minimal exploratory analysis or visualization; only lightweight printouts of small DataFrame heads and no interpretive commentary or visual storytelling.
Evidence
ai-market-intelligence-pipeline/transform.py: prints enriched_df[['title', 'category', 'sentiment']].head()
Predictive Modeling
1/10
Building models that predict
No predictive modeling or rigorous model evaluation is present; the code uses an LLM for extraction/enrichment but does not implement baseline-first modeling, cross-validation, calibration or error analysis.
Evidence
ai-market-intelligence-pipeline/transform.py: use of client.models.generate_content to extract structured data but no modelling or evaluation code
Business Insight & Impact
2/10
Turning analysis into business value
Some operational awareness of business impact is visible - results are uploaded to a stakeholder-facing Google Sheet and deduplicated - but there is no explicit framing of business metrics, cost-of-error analysis or prioritization of actionable KPIs.
Evidence
ai-market-intelligence-pipeline/load.py: pushes deduplicated master dataset to Google Sheets and prints dashboard messages
ai-market-intelligence-pipeline/main.py: orchestration to create daily_enriched_data.csv for stakeholders
Reproducibility & Notebook Hygiene
2/10
Clean, repeatable analysis
Basic modular structure (extract/transform/load) and a requirements.txt are present, but there is no environment pinning, tests, data versioning, CI configuration shown, or reproducible deployment artifacts.
Evidence
ai-market-intelligence-pipeline/requirements.txt: lists runtime dependencies
ai-market-intelligence-pipeline/main.py: orchestrates the ETL steps as a standalone script
Expertise
Analytics• Middle
Technologies
LLM Apps
Google GenAI SDK
Pandas• 3y+
LLM
Recommendations
- Harden the pipeline for production: add retries, backoff, structured error handling around API calls, and unit/integration tests for each module.
- Add reproducibility and observability: pin dependency versions, provide an environment config, add logging/metrics, and a simple CI workflow to run the pipeline on schedule.
- Improve AI reliability: validate and schema-check LLM outputs, add fallback rules for malformed JSON, and log confidence or parse-failure rates for downstream auditing.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior AI/ML Engineer
Confidence: Medium LLM Engineer
LLM-focused junior engineer (approx 1-2 years) who builds API-driven ETL pipelines to enrich news data with LLM extraction. The strongest proven skill is constructing an LLM enrichment flow and simple ETL orchestration as evidenced by transform.py calling the Google GenAI client and main.py coordinating extract-transform-load. There is no evidence of custom model training, evaluation infrastructure, experiment tracking, or production-grade MLOps in the public code.
Model Architecture & Training
1/10
How well models are designed and trained
No custom model architecture or training code; only API-driven calls to a hosted LLM are present.
Evidence
ai-market-intelligence-pipeline/transform.py: client.models.generate_content(...) call to gemini-2.5-flash
Data Pipeline & Feature Engineering
3/10
How data is prepared for models
Basic ETL pipeline: RSS extraction, light cleaning, deduplication and CSV/Google Sheet I/O are implemented but lack robust validation and scaling features.
Evidence
ai-market-intelligence-pipeline/extract.py: fetch_unstructured_data(rss_url) builds DataFrame from feedparser entries
ai-market-intelligence-pipeline/transform.py: published_date cleaning and apply of extract_structured_data over rows
ai-market-intelligence-pipeline/load.py: push_to_google_sheets uses pd.concat and drop_duplicates(subset=['link'])
Experimentation & Evaluation
How results are measured and tested
Not evidenced in public code
MLOps & Deployment
2/10
How models are shipped to production
Simple script orchestration and a basic load step to Google Sheets; no CI/CD, monitoring, or deployment primitives in the code itself.
Evidence
ai-market-intelligence-pipeline/main.py: run_pipeline orchestration calling extract/transform/load
ai-market-intelligence-pipeline/load.py: push_to_google_sheets handles spreadsheet I/O and a SpreadsheetNotFound exception
Computational Efficiency
1/10
How efficiently computing resources are used
No computational optimization for model inference or training is present; only ad-hoc rate limiting via time.sleep is used.
Evidence
ai-market-intelligence-pipeline/transform.py: time.sleep(2) used to pace API calls
Research Depth & Innovation
1/10
Depth of research and new ideas
No research contributions or paper implementations; minimal prompt engineering (JSON-schema prompting) is present but no ablations, baselines or novel algorithms.
Evidence
ai-market-intelligence-pipeline/transform.py: structured JSON prompt string enforcing output schema
Expertise
LLM• Junior
Industries
Information Technology• Junior
Technologies
Python• Senior • 3y+
SQL• Senior • 3y+
PostgreSQL
Snowflake
Claude
GCP
LookML
Beautiful Soup
CI/CD
NumPy• 3y+
Git
Gemini
Google BigQuery
SLI/SLO/SLA
Recommendations
- Develop small LLM-based data enrichment or extraction pipelines where hosted LLMs are used as analyzers and outputs are validated and schema-checked.
- Harden the pipeline for production: add robust LLM-output schema validation, retries/backoff, secrets management (no local credentials.json), and error logging/monitoring.
- Expand into simple MLOps tasks such as adding unit tests, experiment tracking (W&B or MLflow), and CI/CD scheduling to make the pipeline reproducible and maintainable.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: