Skip to content
View KaranAnchan's full-sized avatar
👽
👽

Sponsoring

@fenrir2608

Highlights

  • Pro

Block or report KaranAnchan

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.

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
KaranAnchan/README.md
Karan Anchan — AI Researcher & Engineer

Karan Anchan

M.Sc. Computer Science (AI)  ·  University of Freiburg Freiburg im Breisgau, Germany

Portfolio LinkedIn Email CV


About

I build end-to-end AI systems — from volumetric medical-image segmentation with vision transformers, to Transformer architectures for low-resource translation, to retrieval-augmented generation pipelines in production. I care about measured metrics, reproducible pipelines, and systems that actually ship.


Tech Stack

Tech stack icons

Specialized tooling  ·  Transformers · MONAI · LangChain · MLflow · ONNX · ChromaDB


Experience

Machine Learning Intern  ·  WiZdom Ed

Mangalore, India · Oct 2023 — Oct 2024

  • Built a RAG-based search system using LangChain and ChromaDB for internal document retrieval.
  • Improved data ingestion speed by 40% using recursive text splitting.
  • Developed a cosine-similarity feedback loop to monitor retrieval quality, reaching 90% answer accuracy.
  • Processed and managed a database of 5,000+ educational documents for automated student support.

Featured Projects

PyTorch · MONAI · Vision Transformers

3D segmentation pipeline using UNETR on the BTCV dataset. HU-scaling on DICOM/NIFTI files, Dice-Cross Entropy loss for class imbalance. Result: Mean Dice 0.82 across 13 organs.

PyTorch · Transformers · Gradio

43M-parameter Transformer built from scratch (no nn.Transformer) on 500k AI4Bharat Samanantar pairs. Byte-level BPE, Noam schedule, length-normalized beam search, Gradio demo with cross-attention heatmaps. Result: BLEU 16.93 · chrF++ 41.58 · TER 71.41 (beam-4, held-out 5k test).

➡️ Full portfolio at karananchan.github.io


Education

M.Sc. Computer Science (Artificial Intelligence)  ·  Apr 2025 — Present Albert Ludwig University of Freiburg Focus: Deep Learning · Probabilistic Graphical Models · Statistical Pattern Recognition · Robot Mechanics

B.E. Computer Science  ·  Sep 2020 — Aug 2024 N.M.A.M Institute of Technology, Karkala, India GPA 9.33 / 10  ·  German equivalent 1.3


Certifications

Credential Issuer
Generative AI Automation Vanderbilt University
Reinforcement Learning Specialization University of Alberta · Alberta Machine Intelligence Institute
Machine Learning Operations (MLOps) Duke University
Machine Learning Specialization Stanford & DeepLearning.AI

Languages

🇬🇧 English (C2)  ·  🇮🇳 Hindi (Native)  ·  🇩🇪 German (A2 → pursuing B1)


Pinned Loading

  1. Unetr_3D_Abdomen_Segmentation Unetr_3D_Abdomen_Segmentation Public

    Abdomen 3D Segmentation Using UNETR: Tool for segmenting abdominal organs using the UNETR model. Combines Transformers with CNNs for precise 3D segmentation.

    Python 2

  2. en-hi-nmt-transformer en-hi-nmt-transformer Public

    "Attention Is All You Need" — implemented from scratch in PyTorch and trained for English → Hindi translation on AI4Bharat Samanantar. Includes byte-level BPE, Noam LR schedule, beam search, and a …

    Python

  3. MDP_Optimal_Grid_Sim MDP_Optimal_Grid_Sim Public

    Explore the Gridworld Simulation 🌍🚀! An agent navigates a 5x5 grid to maximize rewards, using the Value Iteration algorithm 🔄. Visualizations 📊 show optimal paths and value convergence. Dive into d…

    Python

  4. 10_Arm_Testbed 10_Arm_Testbed Public

    Explore the 10-Arm Testbed Simulation! 🎲 Utilize Python to test various ε-greedy strategies in a reinforcement learning environment. Visualize and compare agents' performance as they balance explor…

    Python