Available for Internship · Navi Mumbai
RATISH PATIL
Designing intelligent systems — from LLM pipelines and AI agents to products that ship and scale.
scroll to explore
↓
5+
Shipped Products
// about me
AI Architect.
AI Architect.
Builder. Shipper.
I'm Ratish Patil — a final year AI & Data Science student at SIES Graduate School of Technology, Navi Mumbai. I design and deploy intelligent systems end-to-end: from LLM pipelines and RAG architectures to voice-first AI applications used by real people.
I don't just experiment — I ship. Every project here is live, used, and built with production-grade thinking.
LLMs & RAG Pipelines
Retrieval-augmented generation, prompt engineering, and fine-tuned inference
at scale.
Voice-First AI Products
Multilingual AI assistants with VAPI, real-time translation, 8+ Indian
language support.
Computer Vision Systems
Edge-deployed YOLO-based detection and OCR pipelines running at 30 FPS
on-device.
Cloud-Native AI Backends
Full-stack AI on Firebase, GCP, and Vercel — from API design to real-time
telemetry.
// selected work
What I've Built
KrishiSetu
JAN 2026 — PRESENT
- Custom YOLOv8-cls model trained on 22K+ images to diagnose 24 distinct crop diseases with 98% accuracy (efficient CPU inference).
- Multilingual conversational AI voice assistant enabling hands-free navigation and instant eligibility checks for government schemes.
- Automated OCR pipeline parsing land records (7/12 documents) integrated with an LLM matchmaker and RAG workflow for custom PDF reports.
- Google News API scraper integration for real-time agricultural updates (Taaza Khabar) and offline SMS webhook delivery.
98% CV Accuracy
Taaza Khabar Ingestion
OCR & RAG Pipeline
⚡ KrishiSetu System Architecture
- CV Pipeline: Custom YOLOv8-cls trained on 22K+ dataset (24 crop diseases, 98% CPU accuracy).
- RAG & OCR Engine: Tesseract OCR parses 7/12 land records into LLM matchmaker & automated PDF generator.
- Voice Assistant: Multilingual conversational voice AI with scheme eligibility checking.
- Live Telemetry: Google News API scraper (Taaza Khabar) with offline SMS webhook alerts.
Smart Sight
AUG 2025 — NOV 2025
- Standalone, multimodal Edge-AI navigation system deployed on a 4GB RAM Raspberry Pi 4B with real-time computer vision and local NLP.
- Optimized YOLOv4-tiny and OpenCV pipeline for low-latency obstacle detection alongside Tesseract OCR for text recognition.
- Fully offline local voice assistant running TinyLlama and faster_whisper completely on-device for speech-to-text and intent.
- Engineered CloudCam to automatically sync captured photos directly to the user's Microsoft OneDrive account.
10 FPS Edge
4GB RAM RPi 4B
CloudCam (OneDrive)
⚡ Smart Sight Edge Architecture
- Edge Compute: Raspberry Pi 4B (4GB RAM) running lightweight headless Linux OS.
- Vision Model: YOLOv4-tiny + OpenCV executing 10 FPS obstacle detection.
- Offline Voice & LLM: On-device TinyLlama & faster_whisper (Zero cloud dependency).
- Telemetry Sync: CloudCam background thread uploading snapshots to OneDrive.
Review Insight Navigator
FEB 2025 — APR 2025
- Built an automated web scraper using BeautifulSoup and Selenium to extract product reviews and price history across 30+ Amazon listings.
- Applied NLP-based sentiment classification using Scikit-learn to categorize reviews with 85% accuracy.
- Visualized price trends and sentiment breakdowns via Matplotlib and deployed on Streamlit for end-user exploration.
85% Accuracy
30+ Amazon Listings
Matplotlib Vis
⚡ Review Insight Architecture
- Scraper Pipeline: Selenium & BeautifulSoup extracting 30+ Amazon listings.
- NLP Engine: Scikit-Learn TF-IDF vectorizer with 85% sentiment classification.
- Analytics Dashboard: Streamlit interactive frontend with Matplotlib word clouds.
IoT Vehicle Tracker
MAR 2026 — PRESENT
- Standalone Edge tracking system with real-time GPS telemetry and secure Firebase synchronization.
- High-frequency 5-second GPS refresh cycle, systemd-hardened for headless deployment in rugged environments.
- Optimized LTE network data ingestion pipeline achieving <1s upload latency to Firebase Realtime Database.
<1s Latency
Live Telemetry
Field-Ready
⚡ IoT Tracker Architecture
- GPS Hardware: NEO-M8N GPS module with UART serial interface on Raspberry Pi 4B.
- Service Daemon: Python systemd service with auto-recovery & boot initialization.
- Cloud Telemetry: Firebase Realtime Database with <1s end-to-end sync latency.
// toolkit
My Toolkit
Python
TensorFlow
PyTorch
Scikit-learn
OpenCV
Docker
Git
Jupyter
React
Node.js
FastAPI
Firebase
GCP
MongoDB
PostgreSQL
Vercel
Prompt Engineering
Vector Databases
Agent Frameworks
API Design
Real-time Systems
Edge AI
Data Pipelines
System Design
RAG Architecture
// Recruiter Tools
AI Job Matchmaker
Paste any Job Description (JD) below to run an instant semantic match against Ratish's engineering profile, skills, and shipped projects.
// education & leadership
Background
SIES Graduate School of Technology
B.E. — Artificial Intelligence & Data Science
2023 – 2027 · Navi Mumbai
9.26 CGPA
B.K. Birla College
HSC — Science Stream
2021 – 2023
Higher Secondary
NSS Volunteer — Community Leadership
Co-organized the "DREAM RUN" marathon managing logistics for 1,000+ participants,
promoted Blood Donation Drives, Tree Plantation Drives, and coordinated cultural skits.
// courses & credentials
Courses
Advanced RAG & LLM Architectures
CampusX
Let's Build Something.
"Open to internships, collaborations, and interesting problems."
patilratish369@gmail.com
Send an Email