Hema Ramachandran
AI & Data Science · Data Analytics · Intelligent Systems
Turning Data Into Decisions. Building Intelligent Systems With AI.
Pre-final year AI & Data Science student with hands-on experience in Python, SQL, Power BI, Machine Learning, RAG pipelines, and full-stack AI applications.
About
A quick read on who's behind the dashboard.
I'm an AI & Data Science undergraduate who likes working close to the data — cleaning it, querying it, visualizing it, and eventually teaching systems to reason over it. My toolkit spans SQL and Power BI for analytics, Python and Scikit-learn for machine learning, and LangChain-based RAG pipelines for building applications on top of large language models. Alongside project work, I've co-authored applied research and I care about explaining technical decisions clearly to both engineers and non-technical stakeholders.
- Pre-final year AI & Data Science undergraduate at St. Joseph's Institute of Technology, Chennai
- Builds end-to-end data pipelines — from raw data to SQL analysis to Power BI dashboards
- Designs retrieval-augmented generation systems with LangChain, ChromaDB, and open-source LLMs
- Author of research framework for secure healthcare data transmission
- Comfortable translating ambiguous problems into structured, measurable analysis
Skills
The stack behind the analysis — from spreadsheets to vector databases.
Data Analytics
Turning raw tables into decisions
AI & Machine Learning
Models, retrieval, and language systems
Engineering
Shipping analysis as real software
Impact Dashboard
A live-style readout of project output, technology breadth, and skill mix.
Projects Completed Over Time
Skill Distribution
Technology Usage
Featured Projects
Four projects spanning analytics engineering, retrieval-augmented generation, applied security research, and automation.
Commodity Price Intelligence Dashboard
Automated ETL and Power BI dashboards that surface commodity price anomalies for strategic decisions.
Market teams tracked commodity prices manually across scattered spreadsheets, making it slow to spot anomalies or trends.
Built a Python and Pandas ETL pipeline to clean and consolidate multi-source price data, modeled it in SQL, then designed interactive Power BI dashboards for variance and trend analysis.
Cut manual reporting effort and gave stakeholders a live view of volatile commodities and regional price divergence.
Private RAG Chatbot
A document question-answering assistant built on an open-source retrieval-augmented generation stack.
Reading through long private documents to find specific answers is slow, and most QA tools require sending sensitive data to third-party APIs.
Chunked and embedded documents with Hugging Face sentence embeddings, stored vectors in ChromaDB, and served grounded answers with a FLAN-T5 model through a LangChain retrieval pipeline.
Delivered accurate, source-grounded answers entirely on an open-source, self-hostable stack with no external data exposure.
AI Resume Screening & Job Matching
An AI-powered system that screens resumes and matches candidates to relevant job descriptions using GenAI.
Recruiters often spend significant time manually reviewing resumes and comparing candidate profiles against large numbers of job descriptions.
Built a GenAI-powered pipeline that extracts structured information from resumes, evaluates resume integrity, analyzes candidate profiles, and uses an LLM-based scoring workflow to match candidates with relevant job descriptions.
Automated resume screening and job matching into a structured workflow, reducing manual evaluation effort and providing consistent candidate-job matching insights.
AFISL-IDS
A cross-layer adaptive fuzzy-supervised learning framework for secure healthcare data transmission.
Healthcare data transmission networks need intrusion detection that adapts to shifting traffic patterns without flooding operators with false alerts.
Combined fuzzy inference with supervised learning across network layers to classify traffic and flag intrusions with context-aware thresholds.
Framework accepted for poster presentation at IEEE CHANDICON 2026, demonstrating improved detection reliability for secure healthcare communication.
Commodity Price Intelligence
From raw price feeds to a Power BI dashboard stakeholders check every morning.
How the RAG pipeline works
Every document answer is grounded: nothing gets generated without first being retrieved.
Semantic search
Instead of matching exact keywords, each document chunk is converted into a dense vector that captures its meaning. A query is embedded the same way, so the retriever finds passages that are conceptually related — even if they don't share a single word with the question.
Vector databases
ChromaDB and FAISS index those vectors for fast nearest-neighbor lookup, so retrieving the top-k most relevant chunks stays fast even as the document set grows — keeping every answer grounded in retrieved evidence rather than the model's memory alone.
Published Work
Applied research on secure, adaptive systems for healthcare data transmission.
A Cross-Layer Adaptive Fuzzy-Supervised Learning Based Framework for Secure Healthcare Data Transmission
AFISL-IDS
Focused on secure, adaptive communication for healthcare data networks, combining fuzzy inference with supervised learning to detect intrusions across network layers.
Experience
Where the analysis has been put to work so far.
Data Analytics Intern
Kaashiv Infotech
Worked hands-on with Python for data analysis, cleaning, and reporting — translating raw datasets into structured summaries for internal review.
Certifications
Continuous, self-directed learning across data science and programming fundamentals.
The Joy of Computing Using Python
NPTEL · Elite Silver
Python for Data Science
IBM
Crash Course on Python for Data Science
Coursera
Big Data 101
Cognitive.ai
Java Programming
Great Learning
160 Days of Code
GeeksforGeeks