Azure AI Apps Developer

Leeyaz Mohamed K

AI-native Full-Stack Engineer specializing in building production-grade LLM applications, RAG pipelines, and agentic workflows. Delivering robust, low-latency AI systems focused on accuracy and grounding.

About Me

I am an AI-native Full-Stack Developer with 1.5+ years of experience engineering LLM-powered applications using Python, FastAPI, and Django. My expertise lies in designing scalable RAG architectures and orchestrating stateful agentic workflows leveraging LangChain and LangGraph.

I build intelligent, semantic search systems utilizing Azure AI Search, Azure Speech services, and advanced vector/graph databases including ChromaDB, Pinecone, and Neo4j. I integrate AI-native tools like GitHub Copilot, Cursor, and Claude to rapidly prototype, iterate, and deploy reliable solutions.

Beyond AI, I maintain a strong foundation in modern infrastructure, utilizing Docker, Kubernetes, Terraform, and CI/CD pipelines to ensure scalable, responsible, and low-latency cloud deployments.

Technical Expertise

PythonPython
JavaScriptJavaScript
HTMLHTML5
CSSCSS3
BootstrapBootstrap
MySQLMySQL
DockerDocker
KubernetesKubernetes
DjangoDjango
AzureAzure
Azure OpenAI
AI SEARCH AI Search
AI STUDIO AI Studio
ChatGPTChatGPT
◉ claude Claude.ai
RAG RAG
PythonPython
JavaScriptJavaScript
HTMLHTML5
CSSCSS3
BootstrapBootstrap
MySQLMySQL
DockerDocker
KubernetesKubernetes
DjangoDjango
AzureAzure
Azure OpenAI
AI SEARCH AI Search
AI STUDIO AI Studio
ChatGPTChatGPT
◉ claude Claude.ai
RAG RAG

Azure & Cloud AI

Azure OpenAI Azure AI Search Azure AI Studio Speech-to-Text/TTS Container Apps Vertex AI Vector Search

Generative AI & LLMs

LangChain LangGraph LlamaIndex Hugging Face Haystack Ollama RAG Agentic Workflows

Languages & Frameworks

Python TypeScript Django FastAPI Flask JavaScript HTML/CSS/Bootstrap

Vector & Graph Data

ChromaDB Pinecone Neo4j MongoDB MySQL/MariaDB SQL Server

Infrastructure & DevOps

Docker Kubernetes Terraform Apache Airflow CI/CD Cloud Build

Speech & Multimodal AI

Whisper TorToiSe TTS Coqui TTS Voice Cloning Stable Diffusion XL CLIP-ViT

Zoho Platform

Zoho One Zoho CRM Zoho Creator Zoho Books Deluge Scripting

Work Experience

July 2025 – July 2026

AI Full-Stack Engineer

IMACH.AI

  • Designed a hybrid-retrieval chatbot combining Azure AI Search, Neo4j knowledge graph, and MongoDB metadata lookups, integrated with LangGraph/LangFuse — improving answer accuracy by 35% over standard RAG.
  • Built LLM-powered backend services with FastAPI, applying structured prompt engineering to reduce hallucinations.
  • Engineered hybrid rejection pipeline (semantic search + GPU inference) reducing false positives by 42% at <800ms latency.
  • Optimized LLM inference via dynamic batching and FlashAttention-2, cutting per-token cost by 45-50% and improving GPU throughput ~3x.
  • Delivered a real-time speech-to-speech voice assistant (Whisper, TorToiSe) with sub-800ms latency.
Dec 2024 – June 2025

Software Developer

Altos Technologies

  • Developed and maintained dynamic, database-driven web applications end-to-end using Python and Django.
  • Designed and optimized database models (MySQL/MariaDB) and integrated payment gateways.
  • Implemented secure authentication and responsive cross-browser/mobile-compatible UI.
May 2024 – Nov 2024

Software Developer

Horeca Host (Dubai — Remote, Freelance)

  • Built customized business applications on Zoho One (CRM, Creator, Books) to automate hospitality operations.
  • Wrote Deluge scripts for custom functions, workflow automation, and third-party API integrations.

Featured Projects

Posed.io — AI Imaging Platform

Posed.io — AI Imaging Platform

Led an AI imaging system processing 10M+ images/day using Stable Diffusion + CLIP. Applied ONNX/TFLite quantization at sub-8ms latency on T4 GPUs.

Stable Diffusion CLIP ONNX/TFLite T4 GPU

⚡ Cut manual post-production by 70%, reduced cloud costs by 60%.

Next-Gen Inventory Management with RAG

Next-Gen Inventory Management with RAG

Deployed a RAG chatbot over 100K+ inventory, supplier, and procurement records using OpenAI GPT-4, LangChain, Pinecone, FastAPI, and Docker.

GPT-4 LangChain Pinecone FastAPI

⚡ Reduced supply chain search time by 60% with grounded responses.

VendData — Smart Vending Data Solutions

VendData System

Web-based vending machine monitoring platform for tracking machine performance, inventory, and operational data. Contributed to backend and data management.

Django Real-Time Analytics Data Dashboards
KTCC Gulf Finance Management System

KTCC Finance Management System

A modern, secure finance management platform for KTCC Gulf — handling financial operations, records, and authentication. Built with React, Node.js, MongoDB, and AWS.

React Node.js MongoDB AWS
ASK Logikart — Logistics Platform

ASK Logikart

A logistics and delivery management platform for smarter operations — real-time tracking, seamless shipment management, and data-driven decision making.

Logistics Web Platform Real-Time Tracking

AI Engineering Showcase

Hybrid RAG Architecture

Combining Vector Search with Knowledge Graphs (Neo4j) to build stateful, highly-grounded retrieval architectures that outperform standard RAG approaches by significant margins.

Agentic Workflows

Orchestrating multi-agent systems using LangGraph and LangFuse for observable, step-by-step reasoning, memory retention, and dynamic tool execution.

LLM Inference Optimization

Accelerating model serving using structured pruning, dynamic batching, FlashAttention-2, and AWQ quantization to maximize GPU throughput.

Speech & Multimodal AI

Building real-time speech-to-speech pipelines utilizing Whisper and TorToiSe TTS with sub-800ms latencies for natural conversational voice interfaces.

Hybrid RAG Architecture

Combining Vector Search with Knowledge Graphs (Neo4j) to build stateful, highly-grounded retrieval architectures that outperform standard RAG approaches by significant margins.

Agentic Workflows

Orchestrating multi-agent systems using LangGraph and LangFuse for observable, step-by-step reasoning, memory retention, and dynamic tool execution.

LLM Inference Optimization

Accelerating model serving using structured pruning, dynamic batching, FlashAttention-2, and AWQ quantization to maximize GPU throughput.

Speech & Multimodal AI

Building real-time speech-to-speech pipelines utilizing Whisper and TorToiSe TTS with sub-800ms latencies for natural conversational voice interfaces.

Education

Academic Background

Bachelor of Computer Application (BCA)

The American College, Madurai
2021 – 2024 | GPA: 6.5


Higher Secondary Course (+2)

Mary Ann Matric Hr Sec School, Madurai
2020 – 2021 | 79.00%

Let's Connect

Leeyaz Mohamed K

AI Full-Stack Engineer

leeyazm@gmail.com • +91-8778156877 • LinkedIn

Dubai, UAE