ML & Computer Vision Engineer

Esas Khan

I don't just train models — I turn raw point clouds and pixels into systems that see, understand, and hold up outside the notebook.

Portrait of Esas Khan

whoami

— Software Developer & ML Engineer

I'm Esas Khan, a Software Developer working at the intersection of computer vision, machine learning, and the automation, security, and clarity that make those systems trustworthy. My path has moved between research and production — from point cloud segmentation at Fraunhofer IPM to shipping models end-to-end as part of a two-person startup team — and that back-and-forth shapes how I work now: independent, structured, and hands-on, whether I'm building a training pipeline, labeling a dataset, or tracking down why a model is misbehaving. I care about what actually works, not what merely demos well. To me, technology is never the point in itself — it's a means to find the real signal buried in noisy data, and the real value in a tool, for the people and systems that end up relying on it. What keeps me going is a curiosity that doesn't switch off: I like digging into things nobody bothered to explain well, then explaining them better — and that learning doesn't stop at the last commit, or the last training run.

LinkedIn GitHub
Career Timeline

Where I've Worked

From internships in Karachi to shipping point-cloud models in Freiburg — the roles that got me here.

Software Developer — ML Engineer

Aug 2023 — Present

bimeto GmbH · Freiburg, Germany

  • Worked in a 2-person engineering team at an early-stage startup, owning end-to-end development of point cloud processing, data generation, and model training workflows.
  • Generated labeled datasets from architectural projects and used them to train models for point cloud segmentation and modeling.

Research Assistant (HiWi)

Aug 2022 — Jul 2023

Fraunhofer Institute for Physical Measurement Techniques (IPM) · Freiburg, Germany

  • Research on indoor and outdoor 3D point cloud segmentation and scene completion methods.
  • Master's thesis topic: Semantic Scene Completion.

IT Support

Jul 2022 — Present

Data Center, Albert-Ludwigs-Universität Freiburg · Freiburg, Germany

  • Responsible for software license allocation.
  • Handled email-based support and communication.

Freelance Photographer / Videographer / Video Editor

2015 — Present

Self-employed

  • Produced visual media projects for clients across photography and video.
  • Edited and delivered final content for digital channels.

STEM Teacher

Nov 2019 — May 2020

STEM 360 Pvt. Ltd. · Karachi, Pakistan

  • Worked as a STEM instructor, teaching lego robotics and programming to primary and secondary school students.
  • Taught Science, Technology, Engineering and Math (STEM).

Intern

Jun 2018 — Aug 2018

Orient Energy Systems Pvt. Ltd. · Karachi, Pakistan

  • Internship placement in energy systems operations.
  • Supported customer service, repairs, and maintenance of energy generators.

Intern

Jun 2016 — Aug 2016

Pakistan International Airline · Karachi, Pakistan

  • Internship placement supporting maintenance process documentation.
Technologies

What I can do — and why

I work across the full ML lifecycle — from 3D data and geometry to training, deployment, and the systems around it. Tools chosen for relevance, not trend.

Machine Learning & Deep Learning

Model design, training and evaluation for segmentation, reconstruction and scene understanding.

90%

3D Data & Point Cloud Processing

LiDAR, meshes and point clouds — from raw scan to labeled, model-ready dataset.

85%

MLOps & Deployment

Experiment tracking, containerized training, CI/CD for models that reach production.

70%

Computer Vision

I build vision systems that turn raw images and 3D scans into structured understanding — from architecture to deployed model.

OpenCV | Python | Git

Deep Learning

I design and train models with a focus on efficiency and reproducibility — results that hold up outside the notebook.

Python | PyTorch | TensorFlow | Scikit-learn

3D & Point Clouds

Point clouds, meshes and LiDAR scans are my daily material — from raw capture to labeled, training-ready data.

Open3D | LiDAR | 3D Reconstruction

MLOps & Cloud

I track experiments, containerize training, and get models from notebook to something a team can rely on.

Docker | MLflow | W&B | CI/CD | Azure

Portfolio

Projects that show how I work

Selected projects where I owned the ML, data, or systems work end-to-end.

Point Cloud Segmentation

A production pipeline that turns raw architectural scans into segmented, labeled 3D data — from point cloud to trained model.

Python | Open3D | PyTorch | Docker

Spatial Language Models

Use of spatial language models for understanding and generating programmatic descriptions of 3D scenes.

Python | Open3D | PyTorch | Docker

3D Instance Segmentation

A pipeline for segmenting individual instances in 3D point clouds.

Python | Open3D | PyTorch | Docker

Semantic Scene Completion

Master's research at Fraunhofer IPM on completing occluded indoor and outdoor scenes from partial point cloud data.

Python | PyTorch | LiDAR | 3D Reconstruction

Architectural Dataset Generator

A labeling and generation pipeline that turns architectural project data into training-ready datasets for segmentation models.

Python | Open3D | Point Clouds

Agentic AI Experiments

Applied experiments in reflection, planning, tool use and multi-agent systems with modern LLMs.

Python | LLMs | Agents

Availability

Open to new challenges and collaborations

I'm interested in ambitious ML and computer vision problems — research collaborations, freelance work, or long-term teams working at the intersection of data, 3D and systems. If technical depth and real ownership matter to you, let's talk.

Location

Based in Freiburg im Breisgau

Working from the edge of the Black Forest — open to remote work and on-site collaboration alike.

Click to expand