Featured Projects
Flagship AI Systems Project
AI Job Application Copilot
Full-Stack LLM Workflow Automation Platform
● Built a full-stack AI workflow platform that analyzes job descriptions, compares resume variants, recommends the best-fit resume, generates optimized resume content, and creates cover letters, LinkedIn messages, and cold emails.
● Designed the platform around structured LLM outputs, n8n workflow orchestration, PostgreSQL persistence, prompt governance, truth-policy controls, contact discovery, PDF generation, and reusable application history.
Lane 1 — Applied ML
Telecom Customer Churn Prediction
● Built an end-to-end machine learning pipeline for telecom churn prediction, including data cleaning, EDA, feature preparation, class imbalance handling, model training, and evaluation.
● Improved classification robustness using SMOTE / SMOTEENN for imbalanced data treatment.
● Used Optuna for hyperparameter optimization and built a Streamlit app for business-facing predictions.
2-Stage Loan Approval & Valuation System
● Built a two-stage ML workflow combining loan approval classification and loan value regression.
● Used Scikit-learn Pipelines and ColumnTransformer for structured preprocessing and reproducibility.
● Deployed a Streamlit dashboard for business-facing predictions.
Academic Risk & Engagement Prediction
● Built a predictive analytics system to assess student academic risk and engagement potential.
● Performed EDA and feature engineering using demographic, academic, and behavioral data.
● Delivered an interpretable Streamlit dashboard for early-warning insights.
Lane 2 — GenAI / RAG
Insurance Claim Support AI Agent
● Built a memory-enabled insurance support AI agent using RAG, tool-calling, and persistent conversational context.
● Implemented agentic workflows using LangGraph / LangMem concepts for stateful reasoning and structured task execution.
● Designed a production-style support assistant for complex, information-rich business workflows.
AI-Powered Customer Support Copilot
● Built a retrieval-augmented customer support assistant grounded in domain-specific knowledge.
● Designed a document ingestion and retrieval workflow to improve factual consistency and reduce hallucination risk.
● Created a support-focused AI interface with structured backend logic for scalable Q&A.
LLM Web App with CI/CD Automation
● Built an LLM-powered web application with modern AI app delivery practices.
● Combined application-layer AI functionality with Dockerized workflows and CI/CD automation.
● Demonstrated repeatable AI application engineering beyond simple prototyping.
Lane 3 — AI Systems / MLOps
MLOps Jenkins Shared Library CI/CD Platform
● Built a reusable Jenkins Shared Library in Groovy to standardize CI/CD workflows across applications.
● Automated source checkout, container builds, image publishing, and deployment orchestration.
● Connected CI/CD automation with Docker, Minikube Kubernetes, and GCP VM infrastructure.
Object Detection Pipeline with DVC & FastAPI
● Built an end-to-end computer vision workflow using Faster R-CNN with production-style MLOps practices.
● Implemented DVC pipelines and data/model versioning backed by cloud storage for reproducible ML workflows.
● Exposed the trained model through a FastAPI service and validated inference endpoints with Postman.
PCAP StoryTeller
● Built a Python-based network analysis tool that converts packet capture data into structured, human-readable troubleshooting narratives.
● Designed to reduce manual investigation effort and improve support-team clarity during incident analysis.
● Highlights support-focused automation and systems-oriented problem solving.