HiWi @ RWTH Aachen · Looking for robotics & RL roles

I build robots that see, learn, and act.

Master's student in Robotic Systems Engineering at RWTH Aachen. My background is in deep learning and computer vision; right now I'm putting most of my energy into reinforcement learning and getting robots to work outside of simulation.

See my work Get in touch
  • M.Sc. Robotic Systems Engineering
  • 7 publications
  • 3 patents filed
  • RWTH & ISRO research
01 About

From computer vision
to robot learning.

I build systems that help machines see and make decisions. Before moving to Aachen for my Master's, I spent two years as a research assistant and did a stint at ISRO, where I worked on deep learning for satellite imagery — real, noisy data where mistakes actually matter.

Most of my hands-on work so far is in vision and forecasting: segmentation, tracking, generative models, time series. What I want to do next is reinforcement learning and control — closing the loop from perception to action on real robots.

— working on robots that learn from experience.

0 Publications
5 published · 2 under review
0 Patents filed
+ 1 design certificate
0 Segmentation accuracy
U-Net @ ISRO · 38-Cloud
0 B.Tech CGPA
Top 7 in my department
02 Research

Publications & peer-reviewed research

Seven papers so far, spread across medical AI, remote sensing, forensics and agriculture — from classical machine learning up to transformers and generative models.

  • Effective Heart Disease Prediction Using Machine Learning Techniques

    Algorithms · MDPI

    Published DOI ↗

  • Automated Cervical Cancer Prediction using Ensemble Machine-Learning Approaches

    IEEE ICRASET-2023

    Published DOI ↗

  • Early Prediction of Down Syndrome using Deep Transfer-Learning Approaches

    ICNGCIS'23 · Taylor & Francis

    Published DOI ↗

  • A Novel Approach to Glass Identification using Ensemble Learning for Forensics

    ICNGCIS'23 · Taylor & Francis

    Published DOI ↗

  • Automated Fungi Classification using Deep Learning with Explainable AI

    IEEE · IndiaCom 2025

    Published DOI ↗

  • A Comprehensive & Systematic Study in Crop-Health Prediction using Machine Learning

    Journal of Agriculture & Food Research · Elsevier

    Under review

  • Soil-Quality Analysis & Crop-Yield Prediction using Machine-Learning Algorithms

    Crop Design · Elsevier

    Under review

IP Patents & design certification

202321073458

AI & Machine Learning: The New Frontier in Drug Discovery & Development

Filed · pending
202321076113

Blockchain-Enabled Intelligent Surveillance System with AI & IoT in Healthcare

Filed · pending
202321078056

Intelligent Early Forecasting of Air-Quality & Pollution with an Advanced Alert System

Filed · pending
DESIGN

AI-Enabled Robot for UV-C Disinfection

Certified
03 Projects

Selected projects

A few things I've built end to end — mostly perception and prediction, all measured against proper baselines.

01

Real-Time Object Tracking

YOLOv3 detections fed into Deep SORT to track multiple objects through live video — the kind of perception a mobile robot needs before it can do anything useful.

Real-time inference Deep association metric
  • YOLOv3
  • Deep SORT
  • OpenCV
  • CNN
02

Image Colorization with GANs

A generative adversarial pipeline (ResNet-18 + U-Net) that brings grayscale photos back to color, with clearly higher PSNR than the classical baselines.

High PSNR ResNet-18 + U-Net
  • GAN
  • PyTorch
  • ResNet-18
  • U-Net
03

Handwritten Math-Expression Localization

A fully convolutional U-Net that finds and segments handwritten math expressions in document images, beating SegNet and ResNet baselines.

99.68% training 99.46% validation
  • U-Net
  • FCN
  • Segmentation
04

Game-Theory Stock Forecasting

A forecasting model that mixes news sentiment with game-theoretic biases to capture how markets actually behave — decision-making under uncertainty.

RMSE 5.45 → 3.29 R² 0.89 → 0.99
  • Deep Learning
  • Game Theory
  • NLP

More on GitHub ↗

04 Interactive

Reinforcement learning, live.

A Q-learning agent training right here in your browser — no video, no pre-baked result. Tiles rise and change color with their learned state value, arrows show the greedy policy, and the robot drives the best path it has found so far. Drag to rotate · scroll to zoom.

05 Experience

Where I've worked

  1. Research Assistant (HiWi)

    Oct 2024 – Present

    RWTH Aachen · ANTS — applied ML for multi-modal spectroscopy & industrial material sorting

    • Build classification pipelines that sort 7+ materials — plus finer composition-based sub-classes — from NIR, MIR, FIR and hyperspectral data: 0.91 F1 on live industrial sorting, with material properties read straight from the spectra.
    • Recorded the datasets from scratch: millions of spectra captured across varying conditions and hand-annotated, since nothing off-the-shelf fit these materials.
    • Wrote the team's internal tooling — a custom annotation app that cut labeling time, and a graph-generation tool whose plots go into the group's papers.
    • Prototype in the lab; once it proves out, it moves onto the client's production line.

  2. Deep-Learning Research Intern

    Jan – Mar 2024

    Indian Space Research Organisation (ISRO)

    • Semantic segmentation of Landsat-8 imagery with U-Net, SegFormer and Barlow-Twin U-Net on the 38-Cloud dataset (18,000+ patches) — 97.04% accuracy, holding up even on thin clouds and haze.
    • Self-supervised pretraining with Barlow Twins and transformer backbones — Dice 0.8944, recall pushed to 93.28% (BT-UNet).

  3. Research Assistant

    Aug 2022 – Dec 2024

    Pandit Deendayal Energy University

    • Led ML projects on environmental forecasting (PM2.5 / PM10), crop and soil quality, and groundwater contamination — LSTM, GRU, transformers and VAR on 18,000+ data points.
    • Backed the forecasts with proper statistics: Granger causality, Johansen cointegration and ADF tests.

  4. Full-Stack Developer Intern

    May – Jul 2023

    Unitech Technocrats Pvt. Ltd.

    • Built and maintained a Flask web application, improved the enterprise frontend and hardened authentication through custom API integrations.
06 Toolbox

Skills & stack

01 Languages

  • Python
  • C++
  • C
  • SQL
  • Prolog

02 ML & Deep Learning

  • PyTorch
  • TensorFlow
  • Keras
  • fast.ai
  • Hugging Face
  • scikit-learn
  • NumPy
  • Pandas
  • SciPy

03 Computer Vision

  • OpenCV
  • YOLO
  • Timm
  • Pillow
  • VoTT

04 Big Data

  • Hadoop
  • Spark
  • Apache Kafka

05 Web & Backend

  • Flask
  • Django
  • HTML
  • CSS

06 Visualization & Analysis

  • Matplotlib
  • Seaborn
  • Statistical tests

EDU Education

M.Sc. Robotic Systems EngineeringGPA 2.2
RWTH Aachen · Aachen, Germany
B.Tech Computer Science & EngineeringCGPA 9.83
Pandit Deendayal Energy University · 2020 – 2024

CRED Awards & certificates

  • Merit-cum-Means scholarship — top 7 in my department
  • 1st place, General Knowledge Olympiad (2016 & 2017)
  • Stanford Online — Machine Learning
  • NVIDIA — Deep Learning · Anomaly Detection · Data Engineering
  • IIT Madras — Python for Data Science · Entrepreneurship
  • IIT Roorkee — Product & Brand Management
07 Contact

Let's talk.

I'm currently looking for robotics and RL positions, and I'm always open to research collaborations. Email is the fastest way to reach me.

ghetiat@gmail.com Download CV