INIT IALI ZING NEUR AL NET
0 / 100
AVAILABLE FOR NEW OPPORTUNITIES

JAY THORIYA

>_ โ–ฎ

I architect multi-agent AI platforms and ship LLM-powered systems that turn messy enterprise workflows into clean, automated pipelines. Currently leading AI/ML at Safalsoftcom.

0

Shipped
Projects

0

Years
Experience

0

Enterprise
AI Modules

0

OCR
Precision

SCROLL
[
]
[
]
01 / ABOUT

The mind behind the models

I'm an AI/ML engineer obsessed with the intersection of large language models, retrieval systems, and real production constraints.

I've spent the last few years architecting end-to-end AI platforms โ€” from multi-agent procurement systems and intelligent document processing pipelines to generative storybook creators that ship with print-on-demand fulfillment built in.

What drives me: turning messy, manual workflows into automated systems that actually feel intelligent. I write Python, ship on AWS, build with FastAPI, and live in OpenSearch indexes and vector stores.

LLM Architecture RAG Pipelines Multi-Agent Systems OCR + IDP AWS Cloud
jay@portfolio:~

$ whoami

jay_thoriya

$ cat role.txt

Lead AI/ML Engineer @ Safalsoftcom

$ cat location.txt

Ahmedabad, Gujarat, India ๐Ÿ‡ฎ๐Ÿ‡ณ

$ python -c "import skills; skills.list()"

['LLMs', 'GenAI', 'RAG', 'PyTorch', 'TensorFlow',

 'FastAPI', 'AWS', 'Docker', 'OpenSearch', 'OCR']

$ echo $STATUS

> OPEN_TO_OPPORTUNITIES

$ โ–ˆ

What I build

/01

Generative AI & LLMs

Multi-agent architectures and LLM-powered platforms for enterprise automation, intelligent document processing, and agentic workflows.

  • OpenAI
  • Anthropic
  • LangChain
  • RAG
/02

Machine & Deep Learning

ML/DL pipelines for OCR, NLP, computer vision, and predictive analytics โ€” designed to scale with real production data, not benchmarks.

  • PyTorch
  • TensorFlow
  • scikit-learn
  • Pandas
/03

Cloud & DevOps

Scalable AWS infrastructure, Docker-based environments and CI/CD pipelines that keep deployments zero-downtime and release cycles fast.

  • AWS
  • Docker
  • EC2/S3
  • Bedrock
/04

Document Intelligence

High-volume OCR + ML pipelines extracting structured data from messy real-world documents at 98%+ precision. End-to-end automation.

  • OCR
  • Tesseract
  • Vector DB
  • OpenSearch
02 / TECH STACK

Tools I wield every day

LLMs & Generative AI

0%
  • OpenAI
  • Anthropic
  • Bedrock
  • Hugging Face

Python & FastAPI

0%
  • FastAPI
  • Async
  • Pydantic
  • SQLAlchemy

RAG & LangChain

0%
  • LangChain
  • LlamaIndex
  • Chroma
  • OpenSearch

ML & Deep Learning

0%
  • scikit-learn
  • XGBoost
  • Pandas
  • NumPy

NLP & Computer Vision

0%
  • OpenCV
  • Tesseract
  • spaCy
  • Transformers

PyTorch & TensorFlow

0%
  • PyTorch
  • TensorFlow
  • Keras
  • CUDA

AWS & Docker

0%
  • EC2 / S3
  • Bedrock
  • Lambda
  • Docker

Databases & Search

0%
  • OpenSearch
  • PostgreSQL
  • MongoDB
  • Vector DBs
PYTHONโ€ข PYTORCHโ€ข TENSORFLOWโ€ข LLMsโ€ข GenAIโ€ข RAGโ€ข LANGCHAINโ€ข OCRโ€ข OPENSEARCHโ€ข FASTAPIโ€ข AWSโ€ข DOCKERโ€ข PYTHONโ€ข PYTORCHโ€ข TENSORFLOWโ€ข LLMsโ€ข GenAIโ€ข RAGโ€ข LANGCHAINโ€ข OCRโ€ข OPENSEARCHโ€ข FASTAPIโ€ข AWSโ€ข DOCKERโ€ข
03 / TIMELINE

Career trajectory

Aug 2025 โ€” Present Current

Lead AI/ML Engineer

Safalsoftcom

Architected and deployed an ecosystem of 8+ enterprise AI modules for a multi-domain procurement platform. Leveraged LLMs, OpenSearch vector databases, and OCR pipelines to automate end-to-end workflows โ€” reducing manual intervention by up to 75%. Orchestrated scalable AWS infrastructure and Docker containerization, accelerating CI/CD release cycles by 40%.

  • LLMs
  • OpenSearch
  • FastAPI
  • AWS
  • Docker
Mar 2025 โ€” Aug 2025

AI/ML Engineer

Pickabook

Built an end-to-end multi-modal Generative AI platform that crafts personalized children's storybooks. Designed image processing architecture extracting facial features from user-uploaded photos and integrated them into consistent cartoon illustrations. Fully automated the e-commerce + print-on-demand workflow.

  • GenAI
  • Image Processing
  • Stable Diffusion
  • Python
Jan 2025 โ€” Mar 2025

AI Data Engineer / Consultant

Yuktii Venture

Built a predictive social media optimizer using ML + Generative AI + SQL. Automated data scraping pipelines across YouTube, Instagram, and LinkedIn โ€” reducing research effort by 70%. Created visualization dashboards that improved post engagement by 30%.

  • ML
  • GenAI
  • SQL
  • Scraping
Jun 2024 โ€” Sep 2024

LLM Engineer Intern

Bitcoding

Fine-tuned LLMs for improved natural language understanding. Developed intelligent applications integrating LangChain and OpenAI APIs. Designed LSTM models for sequence prediction and chatbots. Implemented RAG techniques for accurate, context-aware responses.

  • LangChain
  • OpenAI
  • LSTM
  • RAG
Mar 2024 โ€” Apr 2024

Machine Learning Intern

Mentorness

Spearheaded ML projects optimizing predictive models. Worked extensively on data cleaning, feature engineering, and visualization. Experimented with multiple algorithms to develop tailored solutions for unique datasets.

  • scikit-learn
  • Pandas
  • Feature Eng.
Dec 2022 โ€” Mar 2023

Web Development Intern

Tritone Music

Built dynamic, engaging web experiences using React.js. Collaborated with the team to streamline workflows and ship intuitive interfaces in a fast-paced environment.

  • React
  • JavaScript
  • UI/UX

Education

2021 โ€” 2025

B.Tech in Computer Science Engineering

Specialization: AI & Machine Learning

Parul Institute of Technology, Vadodara

CGPA ยท 7.6
2020 โ€” 2021

Higher Secondary (Science)

AB Higher Sec. School, Navsari

72%
2018 โ€” 2019

Secondary (10th Grade)

Neelkanth Higher Sec. School, Navsari

70%
04 / SELECTED WORK

Things I've built & shipped

SITA AI Platform
View Modules
AI Platform ยท Featured โ†’

SITA โ€” AI Procurement Platform

Multi-agent enterprise AI platform with 8+ modules powering procurement automation across multiple domains.

  • FastAPI
  • LLMs
  • OpenSearch
  • Docker

Curious about the rest? Check out everything on GitHub.

All Repositories
05 / GET IN TOUCH

Let's build something real.

Got an AI/ML project, an interesting problem, or just want to chat about LLMs and RAG architectures? My inbox is open.

AI PLATFORM ยท 8 MODULES

SITA โ€” AI Procurement Platform

SITA is an AI-powered procurement assistant enabling business users to interact with ERP procurement databases via natural language. Built a suite of 8 AI modules covering the full procurement lifecycle.

SITA AI Procurement Assistant (NL2SQL)

Designed a multi-stage AI pipeline โ€” intent interpretation โ†’ schema mapping โ†’ SQL generation โ€” enabling non-technical users to query ERP data in plain English and receive structured tables and analytics.

  • NLP
  • SQL Gen
  • ERP
  • OpenAI

Supplier Onboarding Automation

OCR pipelines to extract GSTIN, PAN, and MSME registration identifiers from scanned documents, with validation rules and structured ERP-ready output โ€” eliminating manual data entry.

  • OCR
  • Regex
  • Document AI

Invoice Processing & Financial Extraction

Parsed Indian tax invoices to extract buyer/seller GSTINs, line-item tables, product descriptions, quantities, and totals. Built validation pipelines comparing computed vs printed totals.

  • Document Parsing
  • OCR
  • Finance AI

Technical Parameter Recommendation Engine

AI engineering assistant that analyses material descriptions in procurement events and recommends relevant technical specifications (load capacity, material grade, certifications) to standardise RFQs.

  • Classification
  • Recommendation
  • NLP

Vendor Bid Analysis & Award Recommendation

Evaluated vendor bids with net-landed-cost calculations, currency normalisation, and L1/L2/L3 ranking logic. Integrated techno-commercial compliance checks for data-driven supplier selection.

  • Bid Analysis
  • Ranking
  • Finance

Supplier Recommendation Engine

Semantic matching models linking materials with past procurement records using vector similarity search. Scoring algorithms incorporating participation history, performance metrics, and compliance data.

  • Vector Search
  • Semantic Matching
  • ML

Terms & Conditions Suggestion System

Automatically generates commercial T&C clauses for procurement events based on item/service type โ€” differentiating goods vs services, with configurable templates for procurement teams.

  • LLM
  • Template Engine
  • Procurement

Procurement Risk Detection & Audit System

Compliance monitoring system detecting irregular patterns (common bidder IPs, low competition). Gathers ERP evidence and generates structured JSON audit questions for procurement governance.

  • Risk Detection
  • Audit AI
  • Compliance