LLM & AI/ML Engineer · Pen, Raigad, Maharashtra.

Building
intelligent
systems.

I fine-tune large language models, build RAG pipelines, and ship real-time computer vision systems — all on constrained consumer hardware. 5 deployed projects. Diploma AI & ML, MSBTE 2026 — 77.59%.

View Projects Get In Touch
5+ Projects
3B Params Tuned
4GB VRAM Budget
<100ms CV Inference
LLM RAG CV NLP ML DL
AI / ML
Neural Architecture
LLM Fine-tuning✦ RAG Pipelines✦ Computer Vision✦ QLoRA · PEFT✦ LangChain✦ ChromaDB✦ MediaPipe✦ Flask · Gradio✦ LLM Fine-tuning✦ RAG Pipelines✦ Computer Vision✦ QLoRA · PEFT✦ LangChain✦ ChromaDB✦ MediaPipe✦ Flask · Gradio✦
01

Why
choose me

I build end-to-end AI systems on constrained hardware. No cloud dependency. No excuses.

01

LLM Fine-tuning on 4GB VRAM

Fine-tuned LLaMA 3.2 3B using QLoRA + PEFT on RTX 3050. 4-bit quantization, LoRA R=8, gradient accumulation=16. No cloud bills.

MedLLM · QLoRA · PEFT
02

Multi-Agent RAG

StockBot fetches live financial data, reasons across agents, and returns structured analysis via conversational interface.

LangChain · Gemini API
03

Real-time CV <100ms

SignTalkAI delivers live sign language translation via CNN + MediaPipe. Threaded pipeline. Zero frame drop.

OpenCV · TensorFlow · MJPEG
04

End-to-End Ownership

Every project ships complete — data prep, training, deployment, interface. No half-finished repositories.

Flask · Gradio · Git
05

Built From Mobile → Now Shipping on RTX 3050

Started this entire journey coding on a phone with no laptop. Now running local LLMs, training CNNs, deploying RAG apps. Resourcefulness is the real skill.

Acer Nitro V · Ryzen 5 6600H · 16GB RAM · RTX 3050 6GB
02

Technical
arsenal

From raw Python to deployed LLM applications.

LLM Engineering

Core Specialization
Fine-tuning (QLoRA/PEFT)
88
Prompt Engineering
92
RAG Pipelines
85
LangChain
83
Python
90

Computer Vision

OpenCVMediaPipeTensorFlowCNN
80

ML & Data

Scikit-learnPandasNumPyEDA
84

Deployment

FlaskGradioDockerLM Studio
87

Vector DBs

ChromaDBFAISSEmbeddings
82
Full Stack
LLaMA 3.2PEFTQLoRA BitsAndBytesHugging FaceGemini API LangChainChromaDBFAISS FlaskGradioOpenCV MediaPipeTensorFlowScikit-learn PandasNumPyGit DockerLM StudioSQL
03

5 projects.
All deployed.

Not toy demos. Real data, working interfaces, documented code.

01 Featured

StockBot

Multi-Agent RAG · Flask

Multi-agent stock research assistant. Fetches live financial data, runs multi-step reasoning across agents, and delivers structured investment insights via a conversational interface.

Automates data fetching, summarization & comparative stock analysis in one query.
LangChainMulti-AgentRAGGemini APIFlask
Code
02 Featured

SignTalkAI

Computer Vision · Real-time

Real-time sign language translation via CNN + MediaPipe. Threaded frame acquisition, decoupled preprocessing from inference, live MJPEG stream. Sub-100ms on consumer GPU.

Sub-100ms inference on RTX 3050 6GB — zero frame drop in live stream.
CNNMediaPipeOpenCVTensorFlowFlask
Code
03

MedLLM

LLM Fine-tuning · Medical

Fine-tuned LLaMA 3.2 3B on medical Q&A using PEFT/QLoRA — 4-bit quant, LoRA R=8 A=16, gradient accumulation=16. Runs entirely on 4GB VRAM. Gradio interface.

3B model matching larger general LLMs on targeted medical domain tasks.
LLaMA 3.2QLoRAPEFTBitsAndBytesGradio
Code
04

Spam Email Classifier

NLP · Classification

End-to-end NLP pipeline using TF-IDF vectorization and ensemble ML models. Optimized for precision — minimizes false positives on legitimate email.

High precision on test set — critical emails never wrongly flagged as spam.
NLPTF-IDFScikit-learnPython
Code
05

Student Marks Predictor

Regression · EDA · Flask

Regression model predicting student academic performance. Full EDA, feature engineering pipeline, Flask prediction interface. Complete ML pipeline from raw data to deployment.

End-to-end pipeline demonstrating full ML project ownership.
RegressionEDAPandasFlaskScikit-learn
Code
04

Academic
background

2024 – 2026

Diploma in AI & Machine Learning

Sheth Shri Otarmal Sheshmal Parmar College · MSBTE

77.59% First Class with Distinction
Deep LearningNLPComputer Vision Big Data AnalyticsDIPMLOps
2020

HSC — Science

Maharashtra State Board

59.23%
Self-Taught

LLM Engineering & Deployment

Hugging Face · LangChain · Project-based

Applied
Fine-tuningRAGVector DBsLocal LLM
05

Let's
work together

Actively looking for LLM Engineer and AI/ML Engineer roles in Navi Mumbai and Pune. Open to Full Time, hybrid and remote.

Email aryanpatil2903@gmail.com
GitHub AryanPatil2903
LinkedIn linkedin.com/in/aryanpatil
Location Pen, Raigad,Maharashtra