Data Analytics • SQL • Python • RAG • LLM Apps • Machine Learning

Data Analytics | SQL | Python | RAG | LLM Apps | Machine Learning

I build AI and data systems that are structured, testable, and production-aware.

With 13+ years of technical systems experience, I approach AI, ML, and analytics with a strong engineering mindset — focusing on reliability, evaluation, performance, and clean data modeling.

My work spans:

• Generative AI & RAG systems
• Machine Learning pipelines
• SQL-based analytics platforms
• Data cleaning & performance modeling

I specialize in turning raw data and human intent into scalable, measurable solutions.

I treat prompt engineering as a system design discipline — not trial and error.

• Prompt engineering (Zero-Shot, Few-Shot, Chain-of-Thought, ReAct)
• Retrieval-Augmented Generation (RAG)
• LLM evaluation & benchmarking
• Guardrails & fallback logic
• Structured prompt validation pipelines
• API-based LLM integrations

A structured case study demonstrating how evaluation loops and guardrails improved LLM reliability and reduced hallucinations by 30–40%.

Tech: Python · OpenAI API · JSON · Regex · Evaluation Frameworks

Built an end-to-end analytics pipeline to model user performance and engagement patterns.

• Designed staging → clean schema
• Developed leaderboard, streak, and rolling 7-day metrics
• Modeled question difficulty (avg / median / negative-rate)
• Optimized queries using indexing

Tech: PostgreSQL · Advanced SQL · Window Functions · KPI Modeling

Designed a complete SQL-based analytics system for transactional e-commerce data.

• Built revenue, AOV, LTV, and retention metrics
• Implemented cohort and trend analysis
• Created reusable reporting views
• Modeled business performance KPIs

Tech: PostgreSQL · SQL · Aggregations · Window Functions

Developed a structured data-cleaning pipeline to transform raw CSV sales data into analytics-ready datasets.

• Implemented staging → clean workflow
• Standardized inconsistent categorical fields
• Recomputed missing/mismatched totals
• Applied deduplication and validation logic

Tech: SQL Server · T-SQL · Data Cleaning · ETL Concepts

Built an end-to-end ML pipeline for churn prediction using telecom data.

• Data cleaning & feature engineering
• Model training & evaluation
• Business-driven retention insights

Tech: Python · Pandas · scikit-learn · Classification

Developed a fraud detection model addressing severe class imbalance.

• Precision/Recall optimization
• Confusion matrix & F1-score evaluation
• False positive minimization strategy

Tech: Python · Pandas · scikit-learn

• Built a rule-based chatbot with TF-IDF & cosine similarity
• Integrated REST API backend
• Implemented preprocessing & lemmatization

Tech: Python · Flask · NLTK · HTML · CSS

Python · SQL

PostgreSQL · SQL Server

LLMs · Prompt Engineering · RAG · OpenAI API · Evaluation Frameworks

scikit-learn · Classification · Clustering · Model Evaluation

Data Modeling · Window Functions · KPI Design · Cohort Analysis · ETL

Git · GitHub · Flask · REST APIs · Debugging

• Advanced RAG pipelines with evaluation scoring
• ML monitoring & drift detection
• SQL performance optimization techniques
• Production-ready AI system design

I’m continuously learning, building, and refining practical AI and data systems.

Let’s connect if you're interested in structured AI engineering, clean data modeling, or analytics-driven system design.

• 🧠 GitHub: https://github.com/sarahsair25 LinkedIn:https://www.linkedin.com/in/sarahsair

⭐ If you find my projects interesting, feel free to explore, fork, or star them!