Skip to content
Botmer International®
Botmer talent profile

Mahrukh Khalil

Machine Learning Engineer

Machine learning engineer specializing in computer vision, NLP, and deployed AI applications. Her work includes exercise tracking, pose estimation, text-to-speech, panoramic image segmentation, UAV localization, and fitness retrieval systems. She builds Python model pipelines and FastAPI services, optimizes inference, deploys with Docker and AWS, and leads small engineering teams.

4+ years
professional experience
PyTorch & YOLO
computer vision and geometry
TensorFlow & Hugging Face
nlp and generative applications
Meet the engineer

AI engineering with production depth

Where Mahrukh adds value

  • Built computer-vision exercise tracking with YOLO and custom pose landmarks, connecting dumbbell detection and posture analysis to a deployed mobile application.
  • Developed mobile text-to-speech with voice cloning, expressive synthesis, and optimized inference, alongside Dockerized FastAPI model services running on AWS infrastructure.
  • Delivered indoor panorama segmentation, image stitching, alignment correction, and large-scale image-processing optimization using TensorFlow, PyTorch, and OpenCV across computer-vision workflows.
  • Researched UAV localization using georeferenced satellite imagery, feature extraction, pose estimation, model validation, and deep-learning-based aerial localization methods for realtime use.
Explore professional experience
Core competencies

AI systems grounded in software engineering

01

Computer vision and geometry

Builds object detection, tracking, segmentation, panorama stitching, pose estimation, and aerial localization systems, using augmentation, custom landmarks, transfer learning, and model optimization for deployed vision applications.

  • PyTorch
  • YOLO
  • OpenCV
  • MediaPipe
  • Detectron2
  • MMDetection
02

NLP and generative applications

Develops text classification, sentiment analysis, speech generation, retrieval, and conversational AI capabilities with language models, cloned-voice synthesis, structured workflows, and application-specific knowledge sources.

  • TensorFlow
  • Hugging Face
  • LangChain
  • LangGraph
  • RAG
  • TTS
03

Model deployment and APIs

Builds and deploys containerized model inference APIs, optimizes latency, and operates cloud-backed services for mobile and application use, with version control and delivery tooling.

  • FastAPI
  • Docker
  • AWS EC2
  • S3
  • SageMaker
  • Git
04

Data and model development

Creates and augments datasets, trains and fine-tunes models, evaluates predictions, builds image-processing pipelines, and leads small teams delivering AI systems against defined project timelines.

  • Python
  • Data pipelines
  • Model evaluation
  • Fine-tuning
  • Annotation pipelines
  • Scikit-learn
Experience

From models to product delivery

April 2023 – PresentEdgeops

Machine Learning Engineer

Develops production vision, speech, and AI systems for mobile applications. Builds dumbbell tracking, pose analysis, voice-cloning speech generation, FastAPI inference services, and AWS deployments while optimizing models and leading a team of three engineers.

  • Fine-tuned YOLO object and pose models, improving detection accuracy from 82% to 91%.
  • Added custom spine landmarks to MediaPipe and posture-evaluation algorithms.
  • Reduced inference latency by 35% and deployed Docker model APIs on AWS EC2 and S3.
Dec 2022 – March 2023Technical University of Munich

PhD Visiting Research Intern

Researched a deep-learning UAV localization system using georeferenced satellite imagery. Implemented feature extraction and pose estimation for realtime aerial localization, validated model behavior, and contributed to an international journal publication through computer-vision and image-processing work.

  • Improved localization accuracy by 18% over traditional baseline methods.
  • Used Python, PyTorch, OpenCV, and image-processing techniques.
April 2021 – November 2022Revolve AI

Machine Learning Engineer

Developed semantic segmentation and panorama-processing systems for indoor imagery. Tuned models and augmentation, implemented stitching and alignment correction, and optimized large-scale image-processing pipelines using Python, TensorFlow, PyTorch, and OpenCV for improved output quality and efficiency.

  • Improved segmentation accuracy from 76% to 88%.
  • Reduced distortion errors by 30% and processing time by 25%.
May 2019 – March 2021Fiverr

AI & Data Science Freelancer

Delivered AI and machine-learning projects across computer vision and NLP for international clients. Managed requirements, training, and deployment, building sentiment analysis, text classification, object detection, and face-recognition systems with Python and deep-learning frameworks.

  • Completed more than 25 AI and ML projects.
  • Built text classification and sentiment models with up to 90% accuracy.
Selected work

AI systems built for practical use

Highlights from professional projects, product delivery, and specialist work.

Mobile exercise monitoring

Powerblock Dumbbell Tracker

Developed detection, classification, and tracking for dumbbells integrated into an Android application. Combined YOLO-based vision models, PyTorch fine-tuning, and application inference pipelines, alongside posture-analysis work that extends MediaPipe landmarks for realtime feedback in fitness applications.

  • PyTorch
  • YOLO
  • OpenCV
  • MediaPipe
Conversational fitness

AI Fitness Instructor using LLMs

Built a fitness assistant using gym and company workout knowledge with retrieval-augmented generation. Integrated LangChain for language-model response generation and LangGraph for conversational workflow structure, supporting personalized workout guidance, exercise instructions, and questions through an application-specific knowledge pipeline.

  • RAG
  • LangChain
  • LangGraph
Mobile speech

TTS model

Developed a mobile text-to-speech model that generates natural-sounding speech from a reference cloned voice. Added expressive synthesis capabilities and optimized the inference pipeline for low-latency mobile use while maintaining output quality and deployment suitability.

  • Python
  • TTS
  • Model optimization
Image processing

Panorama Generation from Multiple Images

Developed image stitching, alignment correction, and semantic segmentation for indoor panoramic imagery captured across mobile devices and cameras. Tuned model architectures, augmentation, and large-scale image-processing pipelines to improve boundaries, reduce distortion, and shorten processing time.

  • Python
  • TensorFlow
  • PyTorch
  • OpenCV
Education & professional focus

A foundation for practical delivery

Education & qualifications

MS Data Science

NUST (SEECS), Islamabad, Pakistan

2019 – 2022

BSc (Hons) Computer Systems Engineering

The Islamia University of Bahawalpur

2015 – 2019

Professional focusVision models in mobile products

Work connects detection, tracking, pose landmarks, posture metrics, inference optimization, and API deployment to exercise-monitoring applications rather than isolated computer-vision model experiments.

Professional focusEnd-to-end applied machine learning

Experience covers datasets, model training, fine-tuning, evaluation, image pipelines, speech generation, cloud serving, optimization, and small-team leadership across vision and language product requirements.

Talk to Botmer

See if Mahrukh fits your AI engineering team

Share your requirements, current challenges, and delivery goals. Botmer will review the brief and help arrange the next step.

Request a talent match