Skip to content
View Prishita01's full-sized avatar
🎯
Focusing
🎯
Focusing
1 follower · 18 following

Highlights

  • Pro

Block or report Prishita01

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
Prishita01/README.md

Prishita Ghanathe Krishna 👋

ML & Computer Vision Engineer · MS Computer Science (AI/ML) @ Case Western Reserve University

I build production ML systems that turn raw visual data into working products — from custom GPU kernels to end-to-end RAG pipelines. Published researcher. STEM OPT through Jan 2029.

🛠 What I Actually Build

  • Computer Vision — 3D reconstruction (PSNR 24.3 dB / SSIM 0.985), SAM/COLMAP/OpenMVS pipelines, monocular depth from smartphone-only input
  • GPU Systems — Custom Triton convolution kernels validated against cuDNN at ~1e-5 numerical correctness on ResNet-50; benchmarked across 50+ models on H100 NVL & L40S
  • LLM / RAG — Production RAG over arXiv with hybrid retrieval (BM25 + dense + RRF), cross-encoder reranking, Llama 3.2 via Ollama, FastAPI + Docker
  • Distributed Training — 72.6× CPU→GPU speedup with PyTorch DDP + SLURM on HPC clusters
  • Responsible AI — 68.7% bias reduction at 2.3% accuracy cost (FairCredit, validated across 3 datasets)

🔥 Pinned Projects

  1. Scholar (arxiv-rag) — Production RAG with hybrid retrieval, cross-encoder reranking, local LLM, FastAPI + Docker
  2. Triton-ResNet50 — Custom GPU convolution kernels in OpenAI Triton, validated end-to-end on ImageNet
  3. ReconStructAI3D — Monocular 3D reconstruction pipeline (no depth sensors)
  4. FairCredit — Responsible AI framework for credit risk
  5. StockNet HPC — Distributed multi-GPU training pipeline

🧰 Stack

Python · PyTorch · OpenAI Triton · CUDA · OpenCV · COLMAP · OpenMVS · SAM FastAPI · Docker · ChromaDB · Sentence-Transformers · Ollama · LangChain SLURM · PyTorch DDP · NVIDIA Nsight Compute · H100 NVL · L40S Pandas · NumPy · Scikit-learn · Hugging Face · AIF360 · SHAP

📚 Published Research

  • Gaze Detection Using Eye Tracking in Market Research — IJIRCCE, Vol. 11(5), 2023
  • AutoML: Streamlit-Powered ML Pipeline Generator — IJIRCCE, Vol. 11(5), 2023

🌐 Reach me

📧 prishita.ghanathe@gmail.com 💼 linkedin.com/in/prishitagk

Currently: Research Assistant @ CWRU (post-graduation, volunteer role) — actively interviewing for ML Engineer / Computer Vision / Applied ML roles starting 2026.

Pinned Loading

  1. arxiv-rag arxiv-rag Public

    Production RAG system for ArXiv papers with hybrid retrieval, cross-encoder reranking, and local LLM generation

    Python 1

  2. Triton-ResNet50 Triton-ResNet50 Public

    Python 1

  3. ReconStructAI3D ReconStructAI3D Public

    Monocular 3D reconstruction pipeline using SAM + COLMAP + OpenMVS — PSNR 24.3 dB, SSIM 0.985, no depth sensors.

    Python 2

  4. Stock-Prediction-GPU Stock-Prediction-GPU Public

    Distributed multi-GPU LSTM/CNN training pipeline on HPC. 72.6× CPU→GPU speedup with PyTorch DDP + SLURM.

    Python

  5. FairCredit FairCredit Public

    Responsible AI framework for credit risk. 68.7% fairness improvement at 2.3% accuracy cost. Validated across 3 datasets

    Jupyter Notebook