Aspiring Data Scientist - ML/DL & NLP Specialist

VALLALA KOUSHIK RAJ

B. Tech in Computer Science Engineering

**************@*****.*** 91-965******* linkedin.com/in/vallala-koushik-raj-aa4033222/ https://github.com/KoushikRaj484 https://huggingface.co/koushikvkr484 Career Objective:

Aspiring Data Science with strong expertise in Machine Learning, Deep Learning, and NLP. Experienced in building Transformer-based models, multi-task classifiers, and real-time computer vision systems. Seeking to leverage data-driven insights and scalable ML solutions to solve real-world business problems. EDUCATION

B.Tech in Computer Science Engineering, KONERU LAKSIIMAIAIH UNIVERSITY CGPA : 8.4 /10

Percentage : 79%

2020 – 2024

GUNTUR

Intermediate XII (SSC), NARAYANA JUNIOR COLLEGE

Marks : 918/1000

Percentage : 91%

2018 – 2020

HEYDERABAD

X (STATE), TRIVANI TALENT HIGII SCHOOL

CGPA : 8.3

Percentage : 78%

2018

BHADRADRI

KOTHAGUDEM

SKILLS

Programming Languages

Python, C, C++, Java

Deep Learning

ANN, CNN, RNN, LSTM, Transformers, Attention

Mechanism, SentencePiece

Generative AI

Large Language Models (LLMs), Transformer

Architecture, Prompt Engineering, Text Generation, Fine-Tuning, Hugging Face, Model Deployment

Data Analysis & Visualization & Database

EDA, Matplotlib, Seaborn, Basic Power BI, SQL

Machine Learning

Supervised Learning, Basic Unsupervised Learning,

Algorithms, Model Evaluation, Cross-Validation,

Hyperparameter Tuning

Computer Vision

OpenCV, MediaPipe

Libraries & Frameworks

NumPy, Pandas, Scikit-Learn, TensorFlow, Keras,

PyTorch

Tools

Git, GitHub, Jupyter Notebook, Streamlit

Projects

English to Telugu Translation System

Technologies: Python, TensorFlow, Keras, NumPy, SentencePiece, Streamlit

•Designed and trained a 6-layer Encoder–Decoder Transformer model on 21Lakhs+ parallel sentence pairs.

•Implemented subword tokenization using SentencePiece to reduce OOV errors.

•Achieved improved translation quality with optimized attention mechanisms.

•Deployed as a real-time inference web application using Streamlit. Link : https://huggingface.co/spaces/koushikvkr484/English_to_Telugu_translator Face Recognition Attendance System

Technologies: Python, OpenCV, MediaPipe, Scikit-Learn

•Extracted 478 facial landmarks using MediaPipe Face Mesh for high-precision representation.

•Engineered normalized landmark features to improve classification robustness.

•Trained Random Forest classifier achieving high recognition accuracy.

•Automated login/logout tracking and work-hour calculation system.

•Reduced manual attendance effort by 100% through biometric verification. Link : https://github.com/KoushikRaj484/Face-Attendance-System-using-Machine-Learning Hand Sign Recognition & Text-to-Speech System

Technologies: Python, OpenCV, MediaPipe Hands, Scikit-Learn, NumPy, Pandas, gTTS

•Developed a real-time hand gesture recognition system using 21-point hand landmark detection with MediaPipe and OpenCV.

•Built a custom dataset (1,200+ samples per class) and engineered normalized 3D landmark features for robust classification.

•Trained a Random Forest model for real-time gesture prediction with noise reduction mechanisms.

•Converted recognized gestures into continuous text and speech output using gTTS for accessibility. Link : github.com/KoushikRaj484/Hand-Sign-Detection-to-Text-to-Voice-Converstion Multilingual Hierarchical Ticket Classification System Technologies: Python, NLTK, Word2Vec, BiLSTM, TensorFlow

•Built an end-to-end NLP pipeline for multilingual ticket classification (English & German).

•Trained 300-dimensional Word2Vec embeddings on domain-specific corpus.

•Designed a shared BiLSTM encoder with multi-output heads (Type, Queue, Tags).

•Implemented multi-task learning to improve classification consistency.

•Applied multi-label binarization and hierarchical dependency modeling. Link : https://github.com/KoushikRaj484/Multilingual_Hierarchical_Ticket_Classification

Contact this candidate