Muhammad Shariq Khan
AI Engineer

"Architecting End-to-End GenAI Systems & Agentic Workflows"

Details

Why choose me?

I architect end-to-end AI systems rather than just training models, bridging the gap between research and production. Specialized in fine-tuning Multimodal Transformers (VLMs), deploying scalable inference pipelines, and building Agentic AI workflows with strong proficiency in MLOps and LLMOps.

features

My Services

Below are my highlighted services for your requirements.

Generative AI & LLMs

Fine-tuning Multimodal Transformers (VLMs), building Agentic AI workflows with autonomous tool calling, and RAG systems using vector databases.

Autonomous Systems

Building perception and control systems for autonomous driving using sensor fusion (Camera + LiDAR), ROS 2 integration, and real-time processing.

Deep Learning

Developing CNNs, Transformers, RNNs, and Physics-Informed Neural Networks for computer vision, NLP, and state estimation tasks.

MLOps & Deployment

Building scalable inference pipelines, containerization with Docker, CI/CD automation, and cloud deployment on AWS/GCP.

Reinforcement Learning

Implementing Deep Q-Networks, Q-Learning algorithms for autonomous agents, traffic control optimization, and decision-making systems.

API & Backend Development

Building FastAPI backends, ETL pipelines, and real-time data processing systems to support production AI applications.

Data Gathering/Labelling

Acquiring data and creating labels is crucial. Therefore, also providing services to create/gather/label raw data wrt to user end-requirements.

Machine Learning

From Regression to Classification to Unsupervised Learning. Built several ML models for forecasting, classification, clustering etc.

Data Science and Analysis

Presenting Analysis and Insights based on project requirements in the form of PPT, Report or Dashboard.

Portfolio Gallery

Browse through some of my featured projects

SmolAgents AI Assistant:
Fully functional tool-augmented AI assistant with autonomous tool-calling, dual-cloud deployment, and Agentic RAG system using Chroma DB.

AI-Driven Predictive Maintenance:
End-to-end predictive system for aircraft turbofan engines predicting RUL with 99.15% accuracy using hybrid CNN-LSTM model.

LLM-based Driver Fatigue Detection:
Embodied AI agent using ROS 2, LLaMA 2, LoRA/PEFT, and RAG for real-time fatigue detection and vehicle intervention.

Robust Vehicle State Estimation:
Hybrid deep learning framework with GRU, Transformers, and PINNs for vehicle state estimation under GPS dropout.

Smart City Traffic Control
Deep Reinforcement Learning system optimizing traffic light phases for congestion reduction using DQN with Experience Replay integrated with SUMO simulator.

Urban Traffic Analysis & Prediction
End-to-end ML pipeline for large-scale traffic flow analysis using UTD19 dataset with custom Neural Network reducing MSE by 61%.

E2E DriveAI: ROS2 Framework
Self-driving car system using deep learning for steering and speed control with Camera + LiDAR sensor fusion and CUDA-optimized processing.

California Housing Price Predictor
Comprehensive ML pipeline for house price prediction with automated feature engineering using Random Forest achieving low error rates.

Mini-CNN Framework: C++ Engine
Lightweight CNN framework built from scratch using C++17 with LeNet-5 architecture and INT8 quantization reducing memory by 75%.

Autonomous Agent Q-Learning
Self-learning autonomous agent for complex grid environments using Standard Q-Learning and DQN with custom "ShariqQuest" environment.

3DGazeNet Extended Gaze Estimation
Non-intrusive Driver Monitoring System (DMS) for assessing driver alertness with 90.5% gaze classification accuracy using ResNet-18.

LangChain MongoDB Chat Application
Conversational AI chat application using LangChain framework with MongoDB for persistent chat history and memory management.

Vision LLM Traffic Analysis with LoRA
Fine-tuned Vision Language Model for traffic scene analysis using LoRA (Low-Rank Adaptation) for efficient parameter updates and deployment.

IMU Complementary Filter Sensor Fusion
Real-time sensor fusion system combining accelerometer and gyroscope data to accurately track 3D orientation angles (roll, pitch, yaw) using complementary filter algorithm.

Industrial Anomaly Detector: Rotating Machinery
Deep learning-based anomaly detection system for rotating machinery identifying faults and degradation patterns using advanced neural networks for predictive maintenance.

Education

🎓

M.Eng. AI Engineering for Autonomous Systems

Technische Hochschule Ingolstadt, Germany

🎓

B.E. Mechanical Engineering

NED University of Engineering and Technology, Pakistan

Languages

🇬🇧 English — Full Professional Proficiency

🇩🇪 German — Full Professional Proficiency

Tech Stack

Technologies and tools I work with

🧠 AI & GenAI Stack

Python 🔥 PyTorch ⬆️ TensorFlow 🟩 CUDA ⚙️ Scikit-Learn K Keras 🤗 HuggingFace 🦜 LangChain 🤖 SmolAgents 👁️ OpenCV 🎮 Gymnasium 🔧 Tool Calling Pygame

⚙️ Engineering & MLOps

⚡ FastAPI Docker Linux 🤖 ROS 2 🔄 CI/CD AWS Google Cloud 🔴 Streamlit 🤗 HF Spaces 📓 Jupyter Bash Git ☸️ Kubernetes

📊 Data & Analytics

🔢 NumPy 🐼 Pandas 📊 Matplotlib 📈 Seaborn 📊 Power BI 🎨 ChromaDB 🍃 MongoDB 🐘 PostgreSQL MySQL SQLite

🛠️ IDEs & Development Tools

GitHub Copilot 🖱️ Cursor VS Code Google Antigravity

💌 Get in Touch

Have a project in mind? Need an AI solution? Let's discuss!

Feel free to reach out for:

🤝 Collaboration opportunities

💼 Project partnerships

🎓 Knowledge sharing sessions

🚀 New opportunities

📧 engr.m.shariqkhan@gmail.com | 📞 +49 178 2141562

📍 Erkner, Brandenburg, Germany

Muhammad Shariq Khan AI Engineer