MA Muhammad Azhar / Karachi, PK
GEN-AI · RAG PIPELINES · MLOPS · CV & NLP

Building intelligent systems that survive production.

▸ AI Engineer

I design and ship deep-learning models, RAG pipelines, and production-grade AI APIs - from satellite imagery for precision agriculture to medical-imaging diagnostics and Roman-Urdu NLP. Currently shipping AI at scale.

Let's build something ↗ See selected work
0
Models deployed
0
Years building AI
0
Certifications
0
Research papers
gpt-4o · streaming
LangGraph agent
pinecone://prod
fastapi:8000 ✓
azhar@ml-prod · ~/agents/main
LIVE
TOKENS / SEC
1284 ↑
LATENCY P95
142ms ↓
ACTIVE MODELS
07
REQUESTS / MIN
3.4k ↑
● model.training loss · 0.0042 ARCH · 3 - 5 - 6 - 5 - 3 epoch 142/∞
●healthy regionus-east-1 buildv3.2.1 uptime142h 38m
SCROLL
Python TensorFlow OpenAI Anthropic LangChain LangGraph Pinecone ChromaDB FastAPI Docker Kubernetes AWS Azure OpenCV Python TensorFlow OpenAI Anthropic LangChain LangGraph Pinecone ChromaDB FastAPI Docker Kubernetes AWS Azure OpenCV
02 / Toolkit

The stack behind production AI

~ 50+ tools across the ML lifecycle

ML & Deep Learning

/ 06
Python TensorFlow Keras Scikit-learn OpenCV Matplotlib

Generative AI & LLMs

/ 08
OpenAI Anthropic DeepSeek LangChain LangGraph LlamaIndex Flowise Langfuse

Vector & Data

/ 04
Pinecone ChromaDB Milvus Airbyte

Infra & MLOps

/ 07
FastAPI Docker Kubernetes AWS Azure Firebase Linux
// CORE_LOOP ● live
Daily orbit
AI
tools_in_motion = 14 updated daily
03 / Track record

From research labs to production APIs

2020 - present · 4 roles
JUL 2024 - PRESENT
● ACTIVE

AI Engineer

Al Nafi

Developing and deploying AI models and APIs that boosted Alnafi's workflow by 10×. Building generative-AI systems, RAG pipelines, and production-grade APIs that serve thousands of users.

PythonLangChainFastAPIOpenAIDockerPinecone
NOV 2020 - PRESENT
● ACTIVE

Founder & AI Lead

TechShoor

Architected and shipped AI models converted into production APIs across agriculture, healthcare and NLP. Designed Python curriculum for PEPC covering fundamentals through OOP for hundreds of students.

PythonTensorFlowFastAPIDockerAzureFirebase
SEP 2023 - JAN 2024

Junior Lecturer

Institute of Business Management (IoBM)

Taught OOP, Data Structures & Algorithms, Computer Networks and Programming Fundamentals. Authored lab manuals and supervised final-year projects introducing students to cutting-edge AI tooling.

PythonJavaC++DSANetworking
FEB 2021 - JUN 2023

Data Analyst

Unique Traders

Built scalable scrapers to harvest customer data from Google Maps, designed and implemented relational databases, and shipped automated marketing pipelines for client outreach.

PythonSeleniumPostgreSQLAutomation
04 / Selected work

Three systems, three industries

CV · Medical AI · NLP
Computer Vision 01 TechShoor Agri

TechShoor Agri - precision agriculture from orbit

Multi-spectralImagery
Field-levelGranularity
Real-timeInference

An AI-powered platform that turns satellite imagery into actionable yield insights for farmers across South Asia.

  • Deep-learning pipeline on satellite imagery for crop-health monitoring
  • Disease detection and land-condition diagnostics with severity mapping
  • Per-field recommendations to lift yield and cut input waste
PythonTensorFlowFastAPIDockerAzureRasterio
visit agri.techshoor.com →
Medical AI 02 Fracture Detection

MediScan - fracture detection on musculoskeletal X-rays

7-classAnatomical regions
< 1sInference
Azure VMProduction stack

A clinical-grade DL pipeline trained on the MURA dataset to flag abnormalities across seven anatomical regions.

  • 7-category classification with TensorFlow + PyTorch ensembles
  • Outperforms baseline models on validation accuracy & F1
  • Deployed to an Azure VM behind nginx - sub-second inference
PythonTensorFlowPyTorchFastAPIDockerAzure
visit mediscan.techshoor.com →
Natural Language 03 Semantic Analysis

Roman-Urdu Sentiment - NLP for an under-served language

CustomCorpus
Roman-UrduRegional language tooling
DjangoWeb app deployed

Custom NLP pipeline that classifies sentiment in Roman Urdu product reviews, closing a real gap in regional language tooling.

  • End-to-end pipeline: tokenize → normalize → classify on custom corpora
  • Trained with NLTK + Scikit-learn on hand-curated review datasets
  • Web app deployed on Django, exposing scores to non-technical users
PythonNLTKScikit-learnDjango
view on github →
05 / Education

Formal training, always learning

2019 - 2026
MS(AI)

Master of Science in Artificial Intelligence

FAST NUCES, Karachi
AUG 2024 - AUG 2026 Specialization · Generative AI & ML systems
BSCS

Bachelor of Science in Computer Science

Sindh Madressatul Islam University, Karachi
AUG 2019 - AUG 2023 · CGPA 3.17 ★ Best Final-Year Project Award
AI Dev

Diploma in Artificial Intelligence

PIAIC - Presidential Initiative for AI & Computing
APR 2019 - AUG 2021 Percentile 90.23
06 / Research

Published & presented

2 papers · peer-reviewed
Journal

Study on Reinforcement Learning and Deep Reinforcement Learning Schemes: A Review

Sir Syed University Research Journal of Engineering & Technology (SSURJET)
M. Azhar, M. Ahmed, K. Kumar, M. K. Niazi, M. Waqas · published
read paper →
Conference

Usage of Deep Learning Models in Agricultural Satellite Imagery - A Review

Global Research Congress 2023 · SMIU Karachi
M. Azhar, M. A. Khuhro, M. K. Niazi · presented
view venue →
AVAILABLE FOR NEW PROJECTS · Q3 2026

Have an AI problem worth solving?

From a single proof-of-concept to a production RAG stack, I'd love to hear about it.

< 24hResponse time
Karachi, PKBased in
Remote · HybridOpen to
FreelanceEngagements
azharthegeek@gmail.com ↗ LinkedIn GitHub
x ·@azharthegeek medium ·@azharthegeek youtube ·@TechShoor