Manoj A A — ML & GenAI Engineer

About

Who I am

I'm a Computer Science graduate from VTU specializing in Machine Learning and Generative AI, building end-to-end AI systems — from raw data ingestion to deployed, real-time applications that solve real business problems.

My experience spans predictive machine learning models, deep learning pipelines with PyTorch, RAG-based knowledge systems using LangChain and ChromaDB, and agentic AI systems that orchestrate multi-step workflows through natural language and tool execution.

I focus on measurable impact — every project I build includes concrete performance improvements, such as improving model accuracy from 73% to 98%, reducing data preparation time by 60%, and lowering response latency by up to 40%. My goal is always reliable, production-ready AI systems.

Currently focused on advancing skills in Agentic AI architectures, Generative AI systems, and building scalable end-to-end AI applications, while actively seeking full-time Machine Learning Engineer or Generative AI Engineer roles.

GenAI & NLP

RAG Pipelines LLM Applications Text-to-SQL Prompt Engineering Conversational AI Semantic Search LangChain ChromaDB Agentic AI

Machine Learning & Deep Learning

PyTorch scikit-learn XGBoost CNN / ResNet Transfer Learning Feature Engineering Hyperparameter Tuning SMOTE / Imbalance Statistical Analysis

Data & Deployment

Python SQL Pandas / NumPy Flask Streamlit FastAPI MySQL / SQLite Jupyter

Experience

Where I've worked

AtliQ Technologies Private Limited

Data Science Intern · Remote

Mar 2026 – May 2026

Built an end-to-end machine learning pipeline to predict customer price range segments using demographic and behavioral data, including data cleaning, preprocessing, and feature engineering.
Engineered meaningful features (customer behavior scores, income–zone interactions, brand-switch indicators) to improve model performance and capture real-world buying patterns.
Trained and evaluated multiple models (XGBoost, Random Forest, LightGBM), applying hyperparameter tuning to select the best-performing model for accurate predictions.
Developed and deployed a Streamlit application for real-time prediction, enabling stakeholders to make data-driven pricing and customer segmentation decisions.
Performed end-to-end retail data analysis using Python (Pandas, Seaborn) by cleaning, transforming, and integrating multi-source datasets (sales, stores, products) to prepare high-quality data for analysis.
Conducted exploratory data analysis and visualization to uncover patterns in sales performance, product pricing, and store distribution across cities and campaigns.
Engineered key business metrics such as Incremental Revenue % and Sales Uplift to evaluate promotion effectiveness and compare campaign performance (e.g., Diwali vs Sankranti).
Derived actionable insights by identifying high-performing products, stores, and promotional strategies, enabling data-driven decision-making and optimization of marketing efforts.

Python Pandas Seaborn Scikit-learn XGBoost Random Forest LightGBM Streamlit

VDT EDU TANTR Ventures Pvt. Ltd

Data Science Intern · Bangalore

Feb 2025 – May 2025

Architected cohort-based regression models by segmenting customers on age and risk profiles, improving insurance premium prediction reliability by 30% and cutting manual validation time by ~2 hours per dataset.
Enhanced data preprocessing pipeline using VIF, advanced encoding, IQR, and Isolation Forest for outlier handling — improving data integrity by 38% and reducing cleaning effort by ~1 hour per cycle.
Developed and tuned XGBoost models with GridSearchCV, boosting prediction accuracy from 73% → 98% while minimizing extreme errors to 2%, ensuring stable, scalable premium estimation.
Designed RESTful APIs with Flask to serve model predictions to a frontend interface, tested endpoints using Postman with CORS configuration for smooth frontend-backend communication.

Python Pandas NumPy Scikit-learn XGBoost Flask Matplotlib

Projects

Things I've built

✦ Agentic AI In Progress

Agentic AI–Powered HR Management System

Building a production-ready Agentic AI HR Assistant designed to automate end-to-end HR operations using natural language. The system enables users to interact conversationally with an AI agent that autonomously executes tasks such as employee onboarding, leave management, meeting scheduling, and ticket resolution, reducing manual HR workload and improving operational efficiency. The assistant leverages Agentic AI architecture, where the LLM intelligently selects and invokes MCP tools to perform multi-step workflows. These tools connect to modular service managers responsible for handling core HR functionalities such as employee records, leave processing, meeting coordination, and ticket tracking. This design allows the AI agent to orchestrate complex business workflows dynamically instead of relying on static rule-based automation. To ensure scalability and maintainability, the system uses schema-driven APIs and service-layer architecture, enabling structured data validation, clean separation of business logic, and reusable components. Each workflow executes through well-defined schemas, ensuring consistency, reliability, and production-grade performance.

MCP Tool Orchestration Workflow Automation Production Architecture

Python Agentic AI MCP Tools Schema-Driven APIs SMTP LLM Orchestration

✦ GenAI

AI-Powered E-commerce Assistant

Multi-pipeline conversational assistant using semantic routing to direct queries to a RAG FAQ engine, a Text-to-SQL product search layer, or a fallback clarification agent. Supports multi-turn memory through session summarization.

↓ 45% data prep time ↓ 35–40% latency

RAG Text-to-SQL ChromaDB LangChain Groq LLM SQLite Streamlit

View project → Live Demo→

✦ GenAI

Real Estate Research Tool

Domain-restricted RAG system for verified real estate journalism. Automates article ingestion, semantic chunking, and dense embedding generation. Responses grounded strictly in retrieved content — no hallucinations, sub-second retrieval.

↑ 30–35% answer accuracy ↓ 60% prep time

RAG LangChain ChromaDB Groq LLM Streamlit

View project → Live Demo→

◈ Deep Learning

Car Damage Detection

End-to-end computer vision pipeline for automated vehicle damage classification using ResNet50 with transfer learning. Layer freezing, dropout tuning, and class-wise evaluation for robust real-world inference.

↑ 28% accuracy ↑ 35% precision/recall ↓ 30 min/vehicle

PyTorch ResNet50 Transfer Learning CNN Streamlit

View project → Live Demo→

◉ Machine Learning

Credit Risk Prediction

Multi-source credit pipeline predicting loan default probability. SMOTE-Tomek for severe class imbalance, Optuna for automated hyperparameter search. Deployed as a real-time Streamlit scoring tool for instant risk assessment.

↑ 26% accuracy ↑ 32% minority recall

XGBoost LightGBM SMOTE-Tomek Optuna Streamlit

View project → Live Demo→

◉ Machine Learning

Health Insurance Premium Prediction

End-to-end ML system estimating insurance premiums using demographic, lifestyle, and medical data. Cohort-based regression models with XGBoost achieved 98% accuracy. Deployed via Streamlit for real-time premium estimation.

73% → 98% accuracy ↑ 30% reliability

XGBoost GridSearchCV Scikit-learn Flask Streamlit

View project → Live Demo→

▲ Statistical Analysis

Credit Card Launch Analysis

A/B testing framework on 50K+ customer records to evaluate a new credit card variant. Z-test and proportion tests with SciPy to measure statistically significant uplift in spending behavior and support data-backed launch decisions.

50K+ records ↑ 66% faster evaluation

Python SciPy Statsmodels A/B Testing Seaborn

View project →

⬡ Python App

Expense Tracking System

Full-stack expense manager with a modular architecture separating a Streamlit UI from a FastAPI backend. CRUD operations, business logic enforcement, Pandas-powered analytics, and comprehensive test coverage for both layers.

↓ 45% tracking effort ↑ 40% data clarity

FastAPI Streamlit MySQL Pandas REST API

View project →

ML &
GenAI
Engineer

Who I am

GenAI & NLP

Machine Learning & Deep Learning

Data & Deployment

Where I've worked

Things I've built

Credentials

Academic background