Rutuj Dhodapkar
Rutuj Dhodapkar
AI & ML Student  ·  Agentic Systems Engineering  ·  Pune, India
Profile

AI & ML student with hands-on experience building on-device LLM systems and agentic pipelines from scratch. Internship background in ML deployment and data engineering. Projects range from edge AI on Android hardware to real-time computer vision — built independently, outside coursework.

Experience
Syntecxhub · Remote
AI Intern
Feb 2026 — Mar 2026
  • Built and integrated AI-driven modules into the core product stack, cutting manual intervention across two key production workflows.
  • Designed model deployment and orchestration flows within production-grade pipelines.
Data Science Intern
Jan 2026 — Feb 2026
  • Engineered end-to-end data pipelines and feature workflows on large-scale structured datasets.
  • Delivered analytics models benchmarked against business KPIs; findings adopted in BI reporting.
V Analytics
Machine Learning Intern
Jun 2024 — Oct 2024
  • Built end-to-end NLP emotion classification system on a 68,000-row real-world text corpus, achieving 92% model accuracy.
  • Deployed trained model to a live web interface via REST API; integrated into client-facing product.
  • Ran full EDA and feature engineering pipeline (Pandas, NumPy); results shipped in client deliverables.
Achievements
1st
National-Level Hackathon Winner
Sanjivni University · 2025
1st
Paper Presentation Winner
2025
2nd
State-Level Hackathon Runner-Up
Jondhale Polytechnic · 2025
Rutuj Dhodapkar
AI & ML Student · rutujdhodapkar.tech
Projects
Autonomous On-Device AI Agent
Edge AI On-Device LLM

Cross-platform autonomous agent (Android & Windows) running locally via Termux and Ollama — zero cloud, zero API cost. Serves Gemma 3 4B with persistent memory across sessions. Implements 8+ tool integrations: file management, hardware control via Termux:API, WhatsApp automation, steganography, WiFi file transfer, and web search.

Gemma 3 4B fully on-device · 8+ integrated tools · Android & Windows
PythonOllamaGemma3-4BTermuxTermux:APIAndroidWindows
Multi-LLM Browser Automation Pipeline
5-Layer Pipeline

Agentic browser automation system with a 5-layer LLM architecture: intent parsing → task planning → orchestration → vision observation → execution. Persistent Playwright sessions, vision LLM observer for page understanding, and NVIDIA API backend. Designed for platforms where conventional APIs are unavailable.

5-layer LLM pipeline · vision observer · persistent browser sessions
PythonPlaywrightNVIDIA APIVision LLMLLaMA
Plant Pathology Classifier
AgriTechCNN

CNN-based image classifier identifying tomato plant disease from leaf imagery. Trained on labeled agricultural dataset for early-stage pathology detection and intervention.

PythonPyTorchCNNImage Classification