Arpan Pramanik — Full-Stack Developer & AI/ML Engineer
ENGINEERING WITH RIGOR, GROUNDED AI & INTENTIONAL DESIGN.
I am a Computer Science Undergraduate specializing in AI/ML at The Neotia University (9.42 CGPA). My focus centers on building reliable web platforms, intelligent retrieval systems, and machine learning models that bridge scientific research with real-world utility.
From authoring PaperLens AI (an academic research co-pilot with hybrid FAISS+BM25 retrieval) to deploying full-stack web platforms and explainable deep learning pipelines, I emphasize clean architecture, high performance, and visual polish.
Full-Stack Web Systems
[01]Architecting end-to-end applications using React, Next.js, FastAPI, Node.js, and PostgreSQL with robust auth & state management.
Grounded AI & RAG
[02]Building intelligent copilots (PaperLens AI) using hybrid retrieval (FAISS + BM25), Groq LLM API, and structured markdown outputs.
Computer Vision & Deep Learning
[03]Developing multi-headed CNNs, Grad-CAM visual explainability maps, and real-time Streamlit inference engines (FruitQ-GradeX).
Academic Excellence
[04]Maintaining a 9.42 / 10 CGPA in B.Tech CSE (AI & ML) at The Neotia University while shipping 10+ production deployments.
Arpan Pramanik
TECHNICAL ECOSYSTEM.
Languages
Formulating efficient algorithms, concurrent data processing, and type-safe backend abstractions across low-level and high-level execution runtimes.
Python
JavaScript
Java
C
SQL
RAI & Intelligent Systems
Architecting deep learning pipelines, training computer vision models, building explainable AI with Grad-CAM, and deploying RAG architectures.
PyTorch
TensorFlow
Keras
OpenCV
Scikit-LearnFull-Stack Engineering
Engineering high-throughput asynchronous microservices, server-rendered frontend architectures, and resilient RESTful APIs for production environments.
React.js
Next.js
Tailwind CSS
Node.js
Express.js
FastAPIData & Cloud Infrastructure
Designing relational and document schemas, managing vector stores for similarity search, automating CI/CD deployments, and orchestrating cloud services.
PostgreSQL
MongoDB
Supabase (pgvector)
AWS (EC2, S3, Lambda)
Docker
GitHub ActionsMLOps & Developer Ecosystem
Maintaining code quality, versioned pipeline deployments, reproducible container environments, and automated testing suites across the software lifecycle.
Git & GitHub
Linux Administration
Postman
Vite
Vercel
JupyterPROFESSIONAL EXPERIENCE & INDUSTRIAL ROLES.
AI Research Intern
Conducted research on lightweight deep learning architectures by optimizing EfficientNet variants for image classification tasks with a focus on model efficiency and inference speed.
- Evaluated architectural variants across the EfficientNet family
- Benchmarked performance on open-source image datasets
- Analyzed trade-offs between accuracy, parameter efficiency, and FLOPs
- Optimized deep learning models for efficient execution
Conducted research on lightweight deep learning architectures by optimizing EfficientNet variants for image classification tasks with a focus on model efficiency and inference speed.
- Evaluated architectural variants across the EfficientNet family
- Benchmarked performance on open-source image datasets
- Analyzed trade-offs between accuracy, parameter efficiency, and FLOPs
- Optimized deep learning models for efficient execution
SELECTED PRODUCTION SYSTEMS & AI PLATFORMS.
VIEW ALL REPOSITORIES ON GITHUB
PaperLens AI
Autonomous full-stack AI research platform for literature analysis, experiment planning, problem ideation, gap detection, citation intelligence, and benchmark discovery.
NeuroVoice: AI Desktop Assistant
Advanced AI-powered voice assistant with Ollama LLM integration for natural conversations and task automation.
CONFERENCE PAPERS & DEEP LEARNING RESEARCH.
This paper presents a novel multi-task deep learning framework for simultaneous fruit classification and quality assessment using a multi-headed Convolutional Neural Network (CNN). The proposed model achieves state-of-the-art performance on a curated dataset of four Indian fruits (apple, banana, guava, and orange) with two quality classes (good and bad), achieving 98% accuracy in fruit classification and 99% accuracy in quality detection.
Multi-headed CNN with shared feature extractor, EfficientNetB3 architecture, Grad-CAM for interpretability
VERIFIED ACADEMIC & TECHNICAL CREDENTIALS.
Cyber Security Training Certificate
Completed one month summer training and internship program in cyber security fundamentals
Foundations and Practical Applications of AWS
Comprehensive training on AWS cloud services, EC2, S3, Lambda, and cloud deployment strategies
Revolutionizing Signal Processing: The Impact of AI and ML
Short-term training program on AI and ML applications in signal processing
Cyber Security Advanced Training
Advanced six-week summer training program covering ethical hacking and security protocols
Flask Development Certification
Intensive two months training in Flask web development, RESTful APIs, and backend systems
Smart India Hackathon (SIH) 2025 — Internal Hackathon
Participated in the Internal Hackathon for Smart India Hackathon 2025 organized at The Neotia University on 19th September 2025, supported by AICTE, Institution's Innovation Council, and MoE's Innovation Cell
Certificate of Merit — Anuranan
Awarded for contributing an article entitled "AI-Driven Smart Waste Management System" for publication in Anuranan (Vol: 5, Issue: 1), recognising creativity, critical thinking, and academic expression