PARTHA GHOSH

PARTHA GHOSH

Machine Learning Engineer
67, Swamibag Lane, Wari, Dhaka, Bangladesh.

About

Highly motivated and results-driven Machine Learning Engineer with a strong foundation in AI/ML model development, optimization, and deployment. Proven ability to design and implement robust solutions for complex problems, including natural language processing, computer vision, and time series forecasting. Experienced in full-stack development, with a passion for leveraging technology to create impactful applications. Eager to contribute to innovative projects and drive advancements in the field of artificial intelligence.

Work

REVE Systems
|

Machine Learning Engineer

Summary

Developed and deployed advanced AI/ML solutions focusing on natural language processing, computer vision, and forecasting to enhance product capabilities and operational efficiency.

Highlights

Fine-tuned Wav2vec 2.0 and Whisper models with PEFT-LoRA for Bengali Speech-to-Text (STT), conducting performance analysis on augmented datasets to identify and address model weaknesses.

Engineered a RAG-based chatbot for REVE CHAT, integrating GraphRag, LightRag, Multi-Modal Rag, Agentic Rag, and LangChain, leveraging hybrid search (BM25 + dense) for enhanced AI chatbot responses.

Optimized vits model for Bengali Text-to-Speech (TTS) to improve synthesis quality and naturalness.

Implemented model compression techniques using Olive, Quantization, and ONNX, significantly improving model loading and inference times for both online and offline services.

Developed an election commission chatbot utilizing BERT and Bi-LSTM architectures, achieving enhanced accuracy in responses.

Integrated face liveness detection for an NTMC biometric authentication project using OpenCV, CNN, and MediaPipe to prevent spoofing.

Developed robust Time Series Models and a comprehensive forecasting pipeline for precise production quantity and sales amount predictions.

Enhanced the frontend of the REVE CHAT dashboard, significantly improving user experience, usability, and interface responsiveness.

Codelab FZC
|

Software Engineer

Summary

Contributed to game development and web application projects, focusing on user experience and backend functionality.

Highlights

Developed and launched a 2D object detection game and a typing game for Android and iOS using Unity, enhancing user engagement and skill development.

Contributed to the development of a comprehensive admin panel for a key project, implementing critical CRUD operations using Laravel.

Education

Shahjalal University of Science & Technology

BSc

Computer Science & Engineering

Grade: 3.35/4.00

Skills

Programming Languages

Python, Java, C, C++, PHP, C#.

Machine Learning & AI

Pytorch, Tensorflow, OpenCV, LLM (Large Language Models), Langchain, LangGraph, Scikit-learn, MLflow, RAG (Retrieval Augmented Generation), Time Series Models, BERT, Bi-LSTM, Wav2vec, Whisper, PEFT-LoRA, Olive, Quantization, ONNX, CNN, MediaPipe.

Web Technologies

Django, JavaScript, React.js, Node.js, Vue.js, HTML, CSS, Laravel, Express.js, Streamlit.

Databases

MySQL, MongoDB, Firebase.

Tools & Platforms

Linux, Shell (Bash), LaTeX, Microsoft Office, Docker, Git, Unity, CI/CD.

Problem Solving

Competitive Programming, Algorithm Design, Data Structures, Codeforces, Leetcode, LightOJ, Vjudge, Stopstalk.

Research Experience

BanSpeech: A Multi-Domain Bangla Speech Recognition Benchmark Toward Robust Performance in Challenging Conditions

Published by

IEEE Xplore

Summary

Assessed a Bangla ASR model on BanSpeech, a 7.2-hour benchmark; it struggled with spontaneous speech and high out-of-vocabulary rates, highlighting the importance of domain selection.

Development of a Large Corpus of Spontaneous Speech for Improved Speech Recognition in Bangladeshi Bangla

Summary

Created a 200-hour Bangladeshi Bangla spontaneous speech corpus and fine-tuned the multilingual wav2vec 2.0 model, comparing it with a reading corpus (SUBAK.KO) to enhance ASR performance.

Projects

SUST Student Management System

Summary

Developed a web-based system to manage academic data for students and faculty.

Automatic Attendance System Using Face Recognition

Summary

Created an automated attendance tracking system using facial recognition technology.

Shoesly - An E-commerce Website

Summary

Built a full-stack e-commerce platform with distinct user roles and transaction management.

Awards

1st Runner up in AI for Bangla 2.0

Awarded By

AI for Bangla

Achieved first runner-up position in a national artificial intelligence competition focused on Bengali language technologies.

Languages

Bengali

Native

English

Proficient

References

Dr. Mohammad Shahidur Rahman

Professor, Department of Computer Science & Engineering, Shahjalal University of Science & Technology, Sylhet. Email: rahmanms@sust.edu

Volunteer

SUST ACM Student Chapter
|

Joint Secretary

Summary

Served as Joint Secretary, contributing to the organization and execution of academic and professional development events.

Highlights

Assisted in coordinating events, workshops, and competitions to promote computer science knowledge and skills among students.

CSE Society, SUST
|

General Secretary, Assistant General Secretary, Executive Member

Summary

Held multiple leadership roles within the Computer Science & Engineering Society, contributing to student welfare and departmental activities.

Highlights

Managed and coordinated various events and initiatives as General Secretary, fostering community engagement.

Supported leadership team in administrative and organizational tasks as Assistant General Secretary and Executive Member.