About
`` We can make it together ``
Experienced
Expertise
- Agentic AI
- LLMs, GEN AI
- Leading Team
- Product Design
- Model Development & Deployment
- Self Discipline
- Problem Solver
Education
The Neotia University
B.Tech in CSE (AI-ML)
CGPA: 9.0 out of 10
Jul 2021- Jul 2025
SBMV(HS)
H.S in Pure Science
Aggregate: 80%
2021
Experience
Currently working as an AI Developer Intern at NTS Nihon Global, contributing to the development of AI-powered features for a multilingual ed-tech platform.
Worked with a Senior Data Scientist and an international intern on the research project “Artificial Intelligence in Microplastic Detection and Pollution Control,” contributing to model development for automated microplastic identification and scalable pollution mitigation strategies.
Developed and deployed advanced GenAI applications using LangChain, RAG, and LLMs with memory integration—building AI tutors, code reviewers, and Flask-based solutions on AWS for real-world production.
Webel (Undertaking Govt.)
AI Developer Intern
CertificateJune 2024 - August 2024
| On Site- Kolkata, WB
Implemented the project 'News Genre Prediction on Headline.
Streamlined project workflows, enhancing overall performance.
Led a team in successfully delivering a complex DL project on time and within the allocated timeslot.
Managed project timelines, reducing delivery times by 60%.
Collaborated with cross-functional teams. Led a team in delivering the project on time and enhancing project success rates by 90%.
Recent Works
Here are some of my favorite projects I have done lately. Feel free to check them out.
-
Medical ChatbotMediBot -
AI-Powered Chat with Multiple PDFs ChatbotAsk-PdfBot -
Context-Aware Chatbot with MemoryMemoraAI -
AI Agent-Powered Reddit Content BotAutoReddBot Agent -
AI-Powered Transcription & Synthetic Media EngineMediaFlow -
ConvLSTM-Based Human Activity RecognitionActionNet -
AI-Powered Desktop Voice AssistantVoxAssist -
NLP-based classifier System for news headlines.News Classification System
MediBot - AI Powered Medical Chatbot
Implemented a medical knowledge retrieval system using Google Gemini 2.0-Flash LLM. Designed a RAG pipeline with Hugging Face embeddings & Pinecone. AWS infrastructure for deployment.
- Tech - Python, LLM, RAG, Pinecone, AWS
Ask-PdfBot - AI Powered Multi-PDF Chatbot
Developed and deployed a high-accuracy multi-PDF chatbot achieving more than 98% answer relevance, tested by 40+ real users for precise document understanding and intelligent querying.
- Tech - Python, LLM, RAG, Faiss, Streamlit, Streamlit_Cloud
MemoraAI - Context-Aware Chatbot with Memory
Built chatbot with memory using ChromaDB & LangChain". Google embedding model for semantic understanding. Implemented a RAG pipeline with LangChain & Pinecone. Hosted using Guardio.
- Tech - Python, Google API, Chroma DB, LangChain, Guardio
"AutoReddBot Agent – AI Agent-Powered Reddit Content Bot
Built automated Reddit bot using Python, Groq AI, and Llama-3-70B, Integrated PRAW API for content posting and scheduling, Implemented error logging and authentication.
- tech - Python, Llama-3, Groq AI, PRAW, Logging
MediaFlow – AI-Powered Transcription & Synthetic Media Engine
Developed pipeline using Whisper and gTTS for speech transcription and synthesis. Automated media processing with FFmpeg and saved outputs in JSON/text. Video synthesis using MoviePy and custom image backgrounds.
- tech - Python, OpenAI Whisper, gTTS, FFmpeg, MoviePy, JSON
ActionNet – ConvLSTM-Based Human Activity Recognition
Built ConvLSTM model using UCF50 dataset to classify 50 actions. Preprocessed video frames using OpenCV and tuned hyperparameters. Achieved 94% accuracy and added a Streamlit dashboard.
- tech - Python, TensorFlow, ConvLSTM, Streamlit, OpenCV
VoxAssist - AI-Powered Desktop Voice Assistant
Built voice-controlled assistant with speech recognition and TTS. Integrated APIs like Wikipedia, SMTP, and camera control with OpenCV.
- tech - Python, pyttsx3, SpeechRecognition, OpenCV, SMTP, Wikipedia API
News Genre Classification System
Developed NLP-based classifier for news headlines using scraped data. Used TF-IDF, CountVectorizer, and ensemble models. Evaluated model performance using classification metrics.
- Tech - Python, NLP, TF-IDF, Count Vectorizer, Ensemble Learning, Web Scraping
What Users Say
Govt. Recognition
For Leading & Delivering an Accurate & Impactful ML Model
“Aritra delivered a high-performing news classification model with over 88% accuracy.
His solution was recognized by senior officials for its real-world relevance and timely execution.”
— Webel(Govt. of WB Undertaking)
Live User Feedback
For Using MediBot (20+ Users Tested)
“MediBot answered my medical queries faster and more accurately than I expected.
The memory feature made the conversation feel natural — like chatting with a real doctor assistant!”
— Early User, MediBot
User(Student) Feedback
For Ask PdfBot
“This bot saved me hours! I used it to understand multiple research PDFs in one place.
Super useful during exam prep — highly recommend to fellow students.”
— User, Ask PdfBot (Multi-PDF Chatbot)
Reddit User Feedback
For AutoReddBot Agent
“Saw a post generated by AutoReddBot — had to dig in! The writing style was natural, engaging, and totally in sync with the subreddit vibe. That’s AI done right.”
— Redditor
Get In Touch
Feel free to reach out — I’m just a message away!
I'm always here to help and connect.