I'm an MCA graduate specializing in Artificial Intelligence & Data Science, focused on building practical AI systems that connect LLMs, data, retrieval, and backend services into usable applications.
I enjoy working on RAG systems, LLM-powered applications, NLP, machine learning, and Python backend development — with an emphasis on building reliable, end-to-end systems rather than isolated experiments.
🌐 Portfolio: sajalraj-portfolio.vercel.app
- LLM-powered applications using Ollama and Gemini
- RAG & Document Intelligence systems with vector and hybrid retrieval
- Natural Language → SQL applications for structured data
- Python backend services and REST APIs using FastAPI
- Full-stack applications with React and PostgreSQL, deployed and containerized with Docker
- Machine Learning systems for classification, regression and anomaly detection
- Data processing & analysis using Pandas, NumPy and SQL
A RAG-based research system that processes documents and answers questions with inline source citations.
Highlights:
- PDF, DOCX, TXT, Markdown, HTML and CSV ingestion
- Semantic retrieval using Qdrant
- PostgreSQL full-text search
- Hybrid retrieval using Reciprocal Rank Fusion (RRF)
- Research-agent endpoint that generates structured reports
- FastAPI backend, Docker Compose setup
Tech: Python · FastAPI · Qdrant · PostgreSQL · RAG · Docker · Alembic
An adaptive interview platform: upload a resume, choose a role track, and get a grounded interview where every question traces back to an item on the resume.
Highlights:
- Resume parsing and signal extraction from PDFs
- Role tracks for AI/ML Engineer, Backend Engineer and Data Scientist
- RAG over a curated knowledge base using vector retrieval
- Multi-dimension rubric scoring of answers
- Interactive analytics dashboard, retry and early-termination flows
Tech: Python · Gemini API · RAG · ChromaDB · Sentence-Transformers · PyMuPDF · Recharts
An LLM-powered system for querying and analyzing support-ticket data using natural language.
Highlights:
- Natural Language → SQL using Ollama / Llama 3.2, running fully locally
- SQL validation: SELECT-only, single statement, destructive keywords blocked
- Detection of stale high-priority tickets and abnormal resolution times
- FastAPI backend and Streamlit interface
Tech: Python · Ollama · LLM · FastAPI · SQLite · Pandas · Streamlit
A full-stack inventory and order system with real-time stock validation, deployed and containerized.
Highlights:
- Row-level locking (
SELECT ... FOR UPDATE) prevents overselling - Database
CHECKconstraints stop stock from going negative - Docker Compose for local setup
- Deployed on Vercel (frontend) and Render (API)
Tech: Python · FastAPI · React · PostgreSQL · Docker
A computer-vision and multimodal AI application combining object detection, image captioning and contextual generation.
Highlights:
- YOLOv8 object detection
- Custom VGG16 + LSTM captioning model trained from scratch
- Beam-search based caption generation
- Salesforce BLIP integration
- Gemini-powered bilingual (English / Hindi) contextual captions
- Streamlit interface
Tech: Python · YOLOv8 · TensorFlow · BLIP · Gemini · Streamlit
An AI-powered adaptive learning platform that generates quizzes and study material, and adjusts difficulty to the learner.
Highlights:
- AI-generated quizzes, PDF summaries and flashcards
- Gemini API integration
- JWT authentication and REST API architecture
- PostgreSQL-backed data storage
Tech: TypeScript · Node.js · Express · React · PostgreSQL · Gemini API
- Hybrid Fake News Detection System — Logistic Regression plus live web verification (Python, Scikit-learn, NLTK, Streamlit)
- Product Demand Forecasting — Linear Regression vs Random Forest on retail sales data
- Analytics case studies: Supply Chain Delivery · E-commerce Pricing · HR Attrition · Customer Churn
- Amazon Price Tracker Bot — my first self-built project (2024)
- Microlearning Generator — flashcards and quizzes from videos, articles or text (2025)
Focus: LLM Integration · Prompt Engineering · RAG · NLP · Generative AI · Transformers
Focus: Classification · Regression · Feature Engineering · Anomaly Detection · Model Evaluation · Object Detection · Image Captioning
Focus: FastAPI · REST APIs · Backend Development · API Integration · Vector Search · Hybrid Retrieval · Full-Text Search
Focus: Data Cleaning · EDA · SQL Querying · Structured Data Analysis · Descriptive Statistics · Jupyter Notebooks
Building LLM- and retrieval-based systems with Python and FastAPI, and open to AI/ML, LLM and backend engineering roles.
- LLM-powered applications
- Retrieval-Augmented Generation (RAG)
- Python backend engineering
- Data-driven AI systems
- AWS Certified Cloud Practitioner
- Tata — Data Visualization Job Simulation (Forage)
- Deloitte Australia — Data Analytics Job Simulation (Forage)
- EY — Technology Risk Job Simulation (Forage)
Building practical AI systems, one project at a time.