Skip to content
View arpanmukherjee38's full-sized avatar
:shipit:
Focusing
:shipit:
Focusing

Highlights

  • Pro

Block or report arpanmukherjee38

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
arpanmukherjee38/README.md

Typing SVG

LinkedIn Email GitHub Portfolio

Profile Views Followers


🧬 About Me

class ArpanMukherjee:
    def __init__(self):
        self.name        = "Arpan Mukherjee"
        self.role        = "Robotics & AI Engineer"
        self.university  = "UEM Kolkata — B.Tech RAI 2024–2028"
        self.cgpa        = "8.0 / 10.0"
        self.location    = "Kolkata, West Bengal 🇮🇳"
        self.languages   = ["Python", "C++", "C", "SQL"]

    @property
    def research_focus(self):
        return [
            "🔬 Mechanistic Interpretability of Transformer Models",
            "🤖 Autonomous Locomotion & Embedded Robotics",
            "🧠 LLM Agents, RAG Architectures & NLP",
        ]

    @property
    def current_research(self):
        return {
            "project"  : "Transformer Internals Visualizer",
            "model"    : "GPT-2 Small (117M parameters)",
            "method"   : "Activation patching · IOI circuit replication",
            "goal"     : ">85% circuit faithfulness on held-out prompts",
        }

    @property
    def highlights(self):
        return [
            "✅ Merged PR → scikit-learn (#32677) — +15% pipeline perf",
            "✅ Bug-fix PR → SecuScan GSSoC '26 (#320)",
            "📄 Peer-reviewed paper presented @ TSSC 2026, UEM Kolkata",
            "🏢 AI/ML Intern — Edunet Foundation × IBM SkillsBuild",
        ]

    def fun_facts(self):
        return "♟️ Chess player | 🧩 Puzzle solver | 🌐 Open source contributor"


🔭 Research Interests & Current Work

🔬 Active Research

  • 🧠 Mechanistic Interpretability — Probing attention circuits in GPT-2 via activation patching & logit-lens analysis
  • 🦾 Embodied AI — Bio-inspired gait algorithms on low-cost embedded hardware
  • 🛡️ AI-Augmented Security — LLM-based vulnerability detection + patch generation

🌱 Learning & Exploring

  • 📐 Algorithmic Foundations of Optimization
  • 🤝 Multi-Agent LLM Systems & RAG Architectures
  • 🛰️ Federated Learning & Differential Privacy
  • ⚙️ Distributed ML Training at Scale

🤝 Open to Collaborate

  • 🔬 AI/ML research & interpretability projects
  • 🤖 Robotics simulations & hardware builds
  • 🌍 Open-source contributions with real-world impact
  • 💬 NLP, computer vision, autonomous systems

💬 Ask Me About

  • 🐍 Python, C, C++ — algorithms & systems
  • 🔬 TransformerLens, activation patching, IOI circuits
  • 🤖 Arduino, ESP32, servo PWM, ESP-NOW
  • 🧠 Deep Learning, RAG, LangGraph agents

🛠️ Tech Stack & Tools

🔬 ML / Deep Learning / Research

PyTorch HuggingFace scikit-learn OpenCV TransformerLens Plotly

🤖 LLM Agents & NLP

LangGraph ChromaDB FAISS Ollama FastAPI Gemini API

🦾 Robotics & Embedded Systems

Arduino ESP32 ROS2 Gazebo Nav2 SLAM

💻 Languages & Systems

Python C++ C SQL Linux Docker Git

📊 Data Science & Visualization

Pandas NumPy Matplotlib Seaborn Power BI Jupyter


🚀 Featured Projects

🧠 Transformer Internals Visualizer

Repo

Research: How do attention heads in GPT-2 Small (117M params) encode grammatical roles like indirect object identification (IOI)?

Activation patching on 500+ IOI sentence structures · Replicating Wang et al. (2022) circuit-finding methodology · Targeting >85% faithfulness on held-out prompts · Layer-by-layer logit-lens dashboard

Python PyTorch TransformerLens GPT-2 Plotly Mechanistic Interpretability

In Progress Research

🦾 4-Legged Quadruped Spider Robot

Repo

Research: Can bio-inspired gait algorithms on low-cost embedded hardware achieve stable multi-terrain locomotion with <5ms wireless latency?

12 MG996R servos · 3 DOF/leg · Custom aluminium chassis · Sub-5ms ESP-NOW command latency validated across 50+ timed trials · Published at TSSC 2026

Arduino ESP32 C++ ESP-NOW Servo PWM Inverse Kinematics

Published Hardware

🤖 Autonomous Mobile Robot Navigation

Repo

Full ROS 2 navigation stack — SLAM-generated maps at <3cm resolution · Nav2 with A* + DWB planners · YOLOv8 obstacle detection · 92% goal-reach rate across 50+ simulation trials

ROS 2 Python Nav2 SLAM Toolbox Gazebo YOLOv8

Active Robotics

🔍 AI-Augmented Code Vulnerability Scanner

Repo

AST traversal detects structural vulnerabilities · LLM-based patch-generation loop · 1,000+ files/min throughput in benchmarked evaluation · Integrated into DevOps CI/CD pipelines

Python AST Parsing FastAPI LLMs Static Analysis CI/CD

Active Security

🧪 Multi-Agent AI Research Assistant

Repo

Multi-agent RAG system using LangGraph + Gemini API · Orchestrates document retrieval, synthesis, and analytical reasoning for academic research workflows

Python LangGraph RAG ChromaDB FAISS Gemini API

Active Agent

📊 Customer Segmentation — IBM SkillsBuild

K-Means clustering on 5,000+ records · Systematically compared k ∈ {3,4,5,6} via silhouette scores · Identified 4 optimal clusters · +12% downstream model accuracy · +18% projected campaign conversion · Interactive Power BI dashboards presented to 50+ participants

Python Scikit-learn Pandas NumPy Power BI K-Means

Completed Industry


🌍 Open Source Contributions

Project Contribution Impact Status
🔬 scikit-learn PR #32677 — Feature-scaling preprocessing optimization +15% pipeline throughput across benchmark datasets Merged
🛡️ SecuScan (GSSoC '26) PR #320 — SQLite variable-limit bug fix + regression tests Fixed silent data corruption; tests approved after maintainer review Confirmed

💡 Open source is where research meets real-world engineering. I contribute to projects I actually use.


📄 Publication & Presentation

Details
Title Bluetooth-Enabled 3D Printed Quadruped Spider Robot on Arduino Nano
Authors S. Banerjee, A. Ghosh, S. Bhadra, R. Saha, P. Das, A. Mukherjee
Venue 1st International Conference on Transformative Social Sciences (TSSC 2026), UEM Kolkata
Date February 2026
Type Peer-reviewed conference paper · Certificate awarded
Contribution Algorithmic gait design, wireless control architecture, hardware fabrication

💼 Experience & Certifications

🏢 Experience

🤖 AI/ML Intern — Edunet Foundation × IBM SkillsBuild Jan 2026 – Feb 2026 · Remote

  • Empirical study on customer segmentation: K-Means on 5,000+ records, comparing k ∈ {3,4,5,6} via silhouette scores → 4 optimal clusters → projected +18% campaign conversion uplift
  • Engineered preprocessing pipeline (imputation, IQR outlier removal, PCA feature engineering) → +12% downstream model accuracy vs unprocessed baseline
  • Deployed Power BI dashboards presenting behavioural insights to 50+ participants

📜 Certifications

Certificate Issuer Year
🧠 Machine Learning Specialization Stanford × DeepLearning.AI 2025
📊 Data Analytics Professional DeepLearning.AI 2025
🤖 Deep Learning Specialization DeepLearning.AI Mar 2026
🤖 ROS 2 Beginners & Navigation The Construct Feb 2026
☁️ OCI 2025 AI Foundations Oracle Nov 2025

🎯 Achievements

🏅 Achievement 📅 Details
scikit-learn Contributor Dec 2025 Merged PR into one of the world's most-used ML libraries — +15% pipeline performance
GSSoC '26 Contributor 2026 Bug-fix PR confirmed for SecuScan; regression tests approved by maintainer
📄 Conference Author Feb 2026 Peer-reviewed paper accepted & presented at TSSC 2026, UEM Kolkata
☁️ Oracle OCI AI Certified Nov 2025 Valid until Nov 2027

📊 GitHub Stats

Activity Graph


🏆 GitHub Trophies

trophy


🤝 Let's Connect & Build Together

"The best research combines rigorous methodology with the courage to ask questions no one has thought to ask yet."

I'm open to collaborating on mechanistic interpretability research, robotics builds, AI agents, open source, and anything that advances science or makes a real-world impact.

LinkedIn Portfolio Email


🐍 Contribution Snake

Snake animation

Views

⚡ Building the Future of Robotics & AI — One Commit at a Time ⚡

Popular repositories Loading

  1. arpanmukherjee38 arpanmukherjee38 Public

    Config files for my GitHub profile.

  2. C-language C-language Public

    C

  3. mangalyaan-game mangalyaan-game Public

    This is the First Mangalyaan Game. 23-09-2025

    JavaScript

  4. scikit-learn scikit-learn Public

    Forked from scikit-learn/scikit-learn

    scikit-learn: machine learning in Python

    Python

  5. safespace-ai-agent safespace-ai-agent Public

    Python

  6. Machine_learning Machine_learning Public

    HTML