Open to internships & collaborations

Manjushwar Khairkar

B.Tech Data Science & Engineering student building ML, geospatial, and RAG/LLM systems.

Jalandhar, Punjab, India
Download Resumemanjushwarpk.ds.23@nitj.ac.in

About / Education

Dr B R Ambedkar National Institute of Technology Jalandhar

2023 – 2027

B.Tech, Data Science and Engineering

Jalandhar, Punjab, India

Experience

Leadlly

June 2026 – July 2026

Research Data Analyst Intern

  • Built web scraping pipelines using BeautifulSoup, Requests, and spaCy to collect and structure data.
  • Leveraged Google Analytics to analyze user behavior and market trends, surfacing engagement patterns and compiling findings into reports used by the product team.
leadlly.in

GSSoC'25

Aug 2025 – Dec 2025

Open Source Project Admin, Wardha Metro Simulation

  • Managed a Metro Simulation project for hometown Wardha, overseeing 70+ PRs with an 85% contributor retention rate, 80+ repository issues, and mentored 40+ contributors.
  • Built a full-stack platform using Flask, FastAPI, SQLAlchemy, and a Random Forest Regressor, integrating ML-based passenger flow prediction with Dijkstra's algorithm-powered route optimization.
ProgramRepo

Projects

Hasdeo Forest Change Monitoring

Geospatial Data & Image Processing

  • Built a vegetation monitoring pipeline from raw satellite imagery through segmentation, applying cloud masking, augmentation, and multispectral transformations (like NDVI) to expand the dataset to 5000 samples.
  • Engineered GLCM texture, spectral, and morphological features; applied Random Forest for vegetation health classification; used QGIS for segmentation; conducted multi-year change detection.
  • Attained 93% accuracy classifying land cover across five distinct classes.
GitHubLive

RAG Codebase Debugger

Natural Language Processing, LLM & AWS

  • Built a Retrieval-Augmented Generation system enabling natural-language Q&A over a C++ codebase, combining structure-aware chunking, Hugging Face embeddings, and a local FAISS vector index.
  • Implemented Tree-sitter-based symbol-level chunking to split C++ source by classes, namespaces, functions, templates, and enums for more precise context retrieval.
  • Integrated a Groq-powered LLM to generate grounded answers from retrieved code context, reducing unsupported/hallucinated responses.
  • Designed a FastAPI backend for retrieval and Q&A, paired with a Streamlit chat frontend for interactive codebase exploration.
GitHubLive

Technical Skills

Languages

  • C
  • C++
  • Python
  • SQL
  • Shell
  • LaTeX

Frameworks / Libraries

  • NumPy
  • PyTorch
  • Scikit-learn
  • OpenCV
  • FastAPI
  • MySQL

Developer Tools

  • Visual Studio Code
  • Xcode
  • Google Colaboratory
  • Git/GitHub
  • Postman
  • Kaggle

Areas of Interest

  • Software Development
  • Data Structures & Algorithms
  • Object-Oriented Programming
  • Artificial Intelligence
  • Machine Learning
  • Deep Learning
  • Data Analytics
  • Probability & Statistics
  • Operating Systems
  • Computer Networks & Security
  • Database Systems
  • Retrieval-Augmented Generation

Extracurricular & Certifications

Sponsorship Head & Band Member

Octaves: Official NITJ Music Club

2025–2027Club

2nd Runner-up, Cyber Security Competition

Conducted by the official Cyber Security Club of NITJ

2025Certificate

Hacktoberfest: Open Source Contributor

Holopin badge

2025Certificate

Neural Network and Deep Learning Course

Coursera

2025Certificate

Contact

Open to internships and collaborations. Reach me at manjushwarofficial@gmail.com.

View resume