Subtain Malik

About Me

Senior Machine Learning Engineer with 5+ years of experience building production AI systems for enterprise clients such as USPS and Airbus. Expertise in LLM and Computer Vision pipelines, processing 100K+ daily requests with 23× faster inference. Specializes in scalable, high-performance ML solutions that drive measurable business impact.

Technical Skills

Languages

Python, C/C++, Javascript, GO, SQL, NoSQL

Technologies

Machine Learning, DeepLearning, PyTorch, Tensorflow, Computer Vision, LLM, HuggingFace, LangChain, LangGraph, TensorRT, Deepstream, Triton Server, Milvus, Numpy, Pandas, Scikit-learn, FastAPI, Kafka, Elasticsearch, Redis, Docker, AWS, Azure

                        5+
                        Years Experience
                    

                        100K+
                        Daily LLM Requests
                    

                        2.3M+
                        Images analyzed per month
                    

Work Experience

Developed and deployed an Agentic AI application with OpenAI, Locally-Hosted DeepSeek-R1, and LangGraph, reducing manual user efforts by 50%.
Engineered and optimized an in-house LLM by fine-tuning LLaMA 3.2-8B, achieving 23x faster inference.
Architected a high-performance AI pipeline processing 100K+ daily requests, resulting in $50,000 annual cost savings.
Built microservice-based AI solutions for computer vision tasks for 2.3M+ images, leading to 65% reduction in manual inspection time.

Jan 2024 - Present

Senior ML Engineer (Team Lead)

Rapidev

Engineered an enterprise-grade analytics chatbot using LangChain and GPT-4, resulting in 40% faster data retrieval.
Architected a centralized AI model platform with automated CI/CD pipelines handling 20+ models.
Optimized QR code detection by designing a multi-threaded microservice, achieving a 10x performance boost.
Accelerated video analytics processing by developing custom NVIDIA DeepStream plugins, boosting FPS by 50%.

Dec 2021 - Dec 2023

Machine Learning Engineer

Darvis

Pioneered an end-to-end small object detection system for military-grade satellite imagery, achieving 95% detection accuracy.
Enhanced satellite imagery analysis by integrating deep learning-based geospatial models, improving monitoring capabilities by 35%.
Built and deployed a dermatological AI diagnostic tool using custom CNN architecture, achieving 86% classification accuracy.

Jun 2019 - Dec 2021

Machine Learning Engineer

CENTAIC

Featured Projects

A-Cube

Multi Agent System

A multi-agent system using LangGraph, Ollama, Vector Database, face and voice recognition to extract user-relevant information and respond to queries accordingly.

LLM LangGraph Vector DB

OWLSENSE

AI-based Social Media Analysis

Developed complete architecture for AI-based social media analysis including LLM, Transcription, Image Captioning, and Multi-lingual Media Analysis.

LLM Image Captioning Transcription

OmniRoom

Video Analytics Optimization

Implemented Gstreamer plugins and deployed segmentation + pose-estimation YOLO models on Deepstream to optimize FPS of computer vision systems.

Deepstream YOLO Gstreamer

SMART

Surveillance Monitoring System

Surveillance-based system with ROI-based tracking and monitoring of sensitive areas and activity recognition.

Computer Vision Activity Recognition ROI Tracking

Auto Annotation Tool

AI-Assisted Data Labeling

Engineered feature matching, interactive segmentation, SAM and YOLO-based auto-annotation tool for efficient data labeling.

SAM YOLO Feature Matching

AIRAF

Aerial Imagery Research & Analysis Framework

Detect small objects in satellite imagery using YOLO/YOLT architecture, achieving 95% detection accuracy for objects under 10 pixels and reducing analysis time by 80%.

Computer Vision YOLO Satellite Imagery

Education & Certifications

Education 🎖️

MS CS&E | Gold Medal

National University of Science and Technology (NUST)

CGPA: 4.0/4.0

Graduation: Dec 2021

Certifications 📋

Tensorflow Certified Developer
HuggingFace Agentic AI Certification
PyCon Speaker
Mathematics for Machine Learning
AI for Medicine

Publications & Contributions 🌎

Publication: Reimagining Application User Interface (UI) Design using Deep Learning Methods: Challenges and Opportunities (2023)

Open Source: TensorRTX, PytorchX, YOLOv7-Pose

Senior Machine Learning Engineer