Skip to content
View KashishMendiratta's full-sized avatar

Block or report KashishMendiratta

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
KashishMendiratta/README.md

Hi, I'm Kashish Mendiratta

I'm an applied machine-learning engineer and M.Sc. Bioinformatics candidate at Saarland University. I build reproducible ML systems for scientific data, from experiment design and model evaluation through APIs, containers, CI/CD, and cloud deployment.

My current thesis work at the Helmholtz Institute for Pharmaceutical Research Saarland focuses on scaling leakage-aware dataset splitting for molecular machine learning.

Selected work

  • AFM Explorer - A tested, cloud-deployed platform for Atomic Force Microscopy analysis, with FastAPI, Streamlit, classical and supervised ML, Docker, AWS, CI/CD, and an MCP tool interface. Live demo
  • Multilingual Spoken Language Identification - Fine-tuned MMS-300M and XLS-R for 22 Indian languages, diagnosing speaker shortcuts and improving top-1 accuracy to 43.4% with targeted augmentation.
  • Seoul Bike Demand Forecasting - Built end-to-end classification and regression pipelines for corrupted temporal data, reaching 0.852 macro F1 with stacked ensembles and 0.371 OOF MSLE with a boosted-tree blend, with SHAP and robustness analysis.
  • Scaling DataSAIL - A balance-aware approximation for leakage-conscious molecular dataset splitting, reducing large-dataset runtime while preserving split quality.
  • Scribble-Supervised Image Segmentation - Classical and deep-learning segmentation from sparse labels, combining GrabCut and a compact U-Net to reach 75.8% mIoU.

Technical focus

Python · PyTorch · scikit-learn · FastAPI · SQL · Docker · AWS · GitHub Actions · MCP

I'm particularly interested in reliable applied ML, scientific ML, model evaluation, and production AI systems.

Connect

Pinned Loading

  1. afm-explorer afm-explorer Public

    Python

  2. Binary_Image_Segmentation Binary_Image_Segmentation Public

    Python

  3. DataSAIL_scaling DataSAIL_scaling Public

    A balance-aware approximation for scaling leakage-aware dataset splitting, evaluated on four molecular ML benchmarks.

    Python

  4. Multilingual_Speech_Language_Identification Multilingual_Speech_Language_Identification Public

    Jupyter Notebook

  5. seoul-bike-demand-forecasting seoul-bike-demand-forecasting Public

    End-to-end Seoul bike-sharing demand forecasting and classification with robust validation, ensembles, and explainability.

    Jupyter Notebook