Manya Megharaj | Software Engineer & Data Science Student

Hi, My name is MANYA MEGHARAJ

Master’s student in Data Science with practical experience as an Application Developer delivering real-world solutions.

3+ Years Experience

2 ML/NLP Projects

25+ Tools & Technologies

About Me

Motivated and results-driven Software Engineer with a strong foundation in Electronics and Communication Engineering and prior industry experience at Accenture. Known for strong problem-solving abilities, adaptability, and a commitment to delivering high-quality solutions in collaborative, fast-paced environments.

Currently pursuing a Master's in Data Science at the University of Europe, I am driven by continuous learning and a strong desire to contribute to organizational success through innovation, analytical thinking, and a positive, growth-oriented mindset.

My Skills

Programming Languages

Java

Python

SQL

TypeScript

Data Analytics & Visualization

Power BI

Pandas

NumPy

Microsoft Excel

ETL

Frameworks & Libraries

Spring Boot

REST APIs

Hugging Face

PyTorch

Cloud & Enterprise Technologies

AWS (EC2, VPC, RDS, IAM, ALB, Auto Scaling)

Cloud Architecture

REST APIs

Apache Kafka

Azure Data Explorer (ADX)

Tools

Git

Jira

Confluence

Postman

VS Code

SSMS

Microsoft Office Suite (Word, PowerPoint)

SharePoint

Concepts

OOPs

Data Structures & Algorithms

Production Support

Functional Testing

Root Cause Analysis

Agile (Scrum)

Work Experience

Custom Software Engineering Analyst

03/2024 – 03/2025

Accenture | Bangalore, India

Project: McDonald's

Developed responsive Angular audit report web applications for McDonald's and integrated RESTful APIs, enhancing functionality and user experience across the platform.
Administered four enterprise systems (PEAK, GBL, PACE, CRCP) supporting restaurant visit management, master data governance, performance reporting, and escalation workflows managing user access, data accuracy, and role-based permissions at scale while collaborating with cross-functional stakeholders.
Mentored junior engineers on technical practices and codebase onboarding, reducing new-hire ramp-up time by 50%.
Delivered accurate design documentation and drove database changes, unit testing, and deployment verification, ensuring error-free implementation and application reliability.

Application Development Associate

12/2021 – 03/2024

Accenture | Bangalore, India

Project: Canadian National Railways

Delivered production support for a Canadian National Railways project, resolving application incidents using SQL, log analysis, and root cause analysis across microservices, Apache Kafka pipelines, and ADX dashboards.
Led incident response during outages — restarting services, coordinating with vendors (TCS, Tideworks), and routing tickets to engineering teams — contributing to a 15% reduction in downtime.
Coordinated deployment activities, release validation, regression testing, and technical documentation for multiple production environments in Brampton, Malport, Vancouver, and Calgary terminals, ensuring stable production operations.
Recognized with the Smart Delivery Award for strong client relationships and reliable delivery, and praised by the project manager for consistently timely ticket resolution.

Graduate Apprenticeship Trainee

01/08/2021 – 20/12/2021

Qspiders | Bangalore, India

Performed web and mobile application testing using Selenium and Appium.
Worked with Oracle DB for SQL-based CRUD operations.
Designed and executed test cases following the STLC, tracking defects using JIRA.

Trainee

27/06/2020 – 27/07/2020

VI Solutions | Bangalore, India

Used LabVIEW for sensor data preprocessing and IoT-based data acquisition.
Built dashboards to visualize data and machine learning outputs.

Projects Hands-on projects spanning NLP, LLM optimization, and applied machine learning.

NLP View on GitHub →

NLP · Legal Tech

Contract Clause Extraction for Operational Risk Review

Fine-tuned Legal-BERT on the CUAD dataset with a retrieval-assisted TF-IDF pipeline, lifting Mean F1 from 0.480 to 0.688 (+43%) while handling 512-token sliding-window chunking and a 1:61 class imbalance.

Python Legal-BERT TF-IDF CUAD

View Poster →

LLM View on GitHub →

LLM · Caching

Reducing Repeated LLM Calls Through Semantic Caching

Built a semantic caching layer with SentenceTransformers, FAISS, and Llama 3 — cutting LLM calls by 91%, improving latency 11×, and reaching a 90.5% cache hit rate at 89.5% answer accuracy.

SentenceTransformers FAISS Llama 3 Python