I'm a final-year B.Tech AI/ML student building end-to-end machine learning systems — from raw data preprocessing to deployed, user-facing applications. My focus sits at the intersection of applied NLP, retrieval-augmented systems, and predictive modeling, with a product-engineering mindset: every project I ship has a live demo, not just a notebook.
Currently interning as an ML Engineer at NAT IT Services Pvt Ltd, building a production RAG pipeline for document question-answering over large-scale PDFs.
Role: AI/ML Engineering Intern
Focus: NLP · RAG Systems · Predictive Modeling
Philosophy: Ship it, measure it, improve it
Open To: AI/ML Internships · Full-time AI/ML Engineer roles · Research collaborationsLanguages
AI / ML Frameworks & Libraries
App Layer & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Natural Language Processing | ●●●●○ | TF-IDF, text classification, fake-review detection pipelines |
| Retrieval-Augmented Generation | ●●●○○ | LangChain + FAISS + HuggingFace embeddings, PDF QA at 500–1000 page scale |
| Predictive Modeling | ●●●●○ | Random Forest, XGBoost on multivariate time-series (IoT sensor data) |
| Local LLM Integration | ●●●○○ | Ollama-based offline inference for voice assistants and QA systems |
| Model Deployment | ●●●●○ | Streamlit apps, packaged executables (PyInstaller) |
🔎 Fake Review Detector — NLP Binary Classifier
Detects fake online reviews by analyzing writing patterns, tone, and repetition using a full NLP pipeline: preprocessing → TF-IDF vectorization → Scikit-learn classification.
| Stack | Scale | Performance | Deployment | Repository |
|---|---|---|---|---|
| Python, NLTK, Scikit-learn | Held-out test set | 89% accuracy | Live on Streamlit | Fake Review Detector |
Built end-to-end from raw text ingestion to a deployed, publicly usable classifier — designed to demonstrate real-world applicability of classical NLP techniques over deep pipelines where interpretability matters.
🎤 AI Interview Simulator — Multi-Signal Scoring Engine
Role-based mock interview platform supporting 5+ job roles, with an automated scoring engine combining TF-IDF cosine similarity, keyword matching, and concept-based scoring.
| Stack | Scale | Performance | Impact | Repository |
|---|---|---|---|---|
| Python, NLP, Streamlit | 5+ job roles | Evaluation time cut from 5 min → <3 sec | Removes manual review bottleneck | Live Demo |
Designed to replace slow, manual interview evaluation with a deterministic, explainable scoring pipeline — deployed as a live, interactive tool rather than a static script.
🎙️ AKIRA — Personal AI Voice Assistant
Python-based, voice-controlled desktop assistant with wake-word activation and full offline AI response capability via a local LLM.
| Stack | Scale | Performance | Security | Repository |
|---|---|---|---|---|
| Python, PyAudio, PyAutoGUI, psutil, Ollama, PyInstaller | 10+ automated tasks | Offline responses in <2 sec | Fully local, no cloud dependency | AKIRA-AI-Voice-Assistant |
Packaged as a standalone executable — covers desktop automation, system monitoring, and local LLM-backed Q&A entirely offline, with no external API dependency.
⚙️ AI-Driven Predictive Maintenance System (In Progress)
Capstone project predicting Remaining Useful Life (RUL) of industrial engines from IoT sensor time-series data.
| Stack | Scale | Models | Evaluation | Repository |
|---|---|---|---|---|
| Python, Pandas, Scikit-learn, XGBoost | 20,000+ engine cycles, 18 sensor features | Random Forest, XGBoost | RMSE, MAE | Private (in progress) |
Feature engineering and model development on multivariate time-series data, moving from raw sensor logs to regression-based failure prediction.
📄 RAG-Based PDF Question Answering System (Internship Project)
Enterprise-style document QA system allowing natural-language queries over large PDFs with source-referenced answers.
| Stack | Scale | Retrieval | Serving | Repository |
|---|---|---|---|---|
| LangChain, FAISS, HuggingFace Embeddings, Ollama, Streamlit | 500–1000 page documents | Vector similarity search | Local LLM inference | Private (NAT IT Services) |
Built during ML internship at NAT IT Services — covers chunking, embedding, vector storage, and retrieval-augmented answer generation with citations.
Machine Learning Intern — NAT IT Services Pvt Ltd Jun 2026 – Present · Hyderabad, India
Building a RAG-based PDF Question Answering System enabling natural-language search over large-scale documents, with source-referenced answers. Also performed Playwright-based web scraping automation for structured extraction of medical conference abstracts.
LangChain FAISS HuggingFace Ollama Streamlit Playwright
| Recognition | Details |
|---|---|
| HackX Hackathon Participant | Jan 2026 — built a practical solution under time constraints |
| Kaggle Housing Prices Competition | Submitted predictions validating Intro to ML coursework |
Kaggle
Workshops & Training
Learning:
- Advanced RAG architectures & vector database optimization
- Production-grade LLM application design
Building:
- AI-Driven Predictive Maintenance System (capstone)
- RAG-based PDF QA System (internship)
Exploring:
- Agentic workflows with LangChain
- Fine-tuning open-source LLMs for domain-specific tasks
Open To:
- AI/ML Engineering roles
- Applied NLP / RAG-focused opportunities
- Research collaborations
"Ship it, measure it, improve it — one deployed system at a time."