👋Hey, I'm Debrup!

I'm an AI Engineer with experience in developing and deploying AI-driven solutions. I've contributed to open-source projects and have a strong background in machine learning and data science.

I enjoy working on innovative projects, exploring new technologies, and continuously learning to enhance my skills.

Debrup Paul

Professional Experience

My journey in AI and Software Development

AI Engineer

OneByZeroFebruary 2025
  • Developed and deployed a real-time AI-driven IVR (Interactive Voice Response) and Voice Agent solution tailored for the ASEAN region over Twilio.
  • Implemented, Deployed, and scaled the Faster Whisper Transcriber Service using WebSockets, migrating from Azure Speech with reduced latency and improved WER through server-side buffering and end-of-speech detection.
  • Integrated real-time noise reduction to enhance transcription accuracy and system performance.
WebSocketsFasterWhisperRayAWS EKSDockerAWS BedrockSileroVADLocust

Google Summer of Code Contributor

C2SI, Score LabsMay 2024 - July 2024
  • Project aims to built Knowledge Graph over Text and Code Modalities over Documentations and perform retrieval for code generation and understanding.
  • Built 4 modules Knowledge Graph builder, Retriever, Experiments, Evaluator handling complete KG-RAG workflow.
  • Optimized retriever LLM's using 8,4 bit Quantization, PEFT methods served in Ollama for local usage.
HuggingFacePyTorchLlama-IndexNeo4jLangChainOllama

Research Intern

Jadavpur UniversitySeptember 2024
  • Conducted research on DeepFake detection in videos, benchmarking against DFDC and Face Forensics++ datasets under Dr. Ram Sarkar's supervision.
  • Investigated various keyframe extraction techniques and implemented feature fusion methods, focusing on both spatial and frequency domain feature extraction.
PyTorchKerasOpenCVDeep LearningComputer Vision

NLP Research Intern

IIIT-Hyderabad, LTRC2024
  • Utilized Knowledge graphs and transformer models for natural language generation from USR. Built an instance based dependency parser for generating Knowledge Graphs.
  • Mentored Sanskrit Scholars on NLP tasks and implementations.
Keras 3.0NLPHuggingFaceTransformersPyTorch

Data Science Intern

Asanify TechnologiesMay 2023 - July 2023
  • Developed GlobalClients database and generated leads through Web Scraping, achieving 15% growth in foreign clients, built visualizations on scraped data using Tableau.
  • Designed and implemented an end-to-end Text-to-SQL pipeline, significantly reducing analysts' workload and enhancing efficiency by automating query generation.
ScrappyAuto-ScraperRegexllama-indexOpenAIPineConeTableau

Education

Academic Background and Achievements

Birla Institute of Technology and Science, Pilani

Integrated BE & MSc in Mathematics
Apr 2021 - May 2026•Goa, India

Relevant Coursework

Computer ProgrammingNumerical MethodsFoundation of Data Science

Advanced Studies

Operation ResearchStatistical InferenceLinear AlgebraProbability and StatisticsAI

Delhi Public School Ruby Park

Higher Secondary (CBSE)
Apr 2008 - Apr 2021•Kolkata, India
Class 12 (PCM):95%
Class 10 (AISSCE):91%

ProjectsAll Projects

Open-source projects I've worked on over the years.
Communication Analysis Platform

Advanced platform for analyzing communication patterns in human-AI driving simulator experiments

FMCG-Insight-AI

Real-time computer vision system for FMCG product detection and freshness prediction

Deepfake Detection Research

Advanced deepfake detection system using spatial-frequency features and adaptive fusion

YojnaPath

Real-time AI-driven IVR and Voice Agent solution for ASEAN region

Knowledge Graph RAG

Knowledge Graph over Text and Code Modalities for Documentation

Text-to-SQL Pipeline

End-to-end Text-to-SQL pipeline for automated query generation

Hate Speech Detection

Real-time hate speech detection using RNN-LSTM architecture

Connect

Let's get in touch!

NotesSee All

Personal notes about life, music, projects, and everything else.

ArticlesSee All

Guides, references, and tutorials.

Deep DivesAll Topics

Long-form tutorials on a variety of development topics.