Skip to content
View Nossks's full-sized avatar

Block or report Nossks

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 supported. This note will be visible to only you.
Report abuse

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

Report abuse
Nossks/README.md

Hey there, I'm Aryan Gairola 👋

🚀 Data Scientist | Competitive Programmer | AI Enthusiast

I love building projects that solve real-world problems, compete in programming contests, and explore the frontiers of AI. Currently, I’m diving deep into, Gen-AI, NLP ,Deep-Learning, and scalable ML systems, Computer Vision.


💻 Tech Stack

  • Programming Languages: Python, C++, C
  • Data Science & ML: Pandas, NumPy, scikit-learn, RAG, Tensorflow, Keras, NLTK, Embedded Systems, Recommendation System
  • Web Development: Flask, HTML, CSS, JS, Jinja2
  • Computer Vision: OpenCV, MediaPipe
  • Databases: Vector-Databases for RAG
  • Tools & Others: Git, VS Code, Jupyter Notebook, Linux

🔥 Projects

1️⃣ FinGuard | Privacy-First Fraud RAG

  • The Build: Developed a RAG pipeline using CyborgDB and LangChain to encrypt vector embeddings, ensuring data privacy in forensic retrieval.
  • Key Innovation: Engineered an intelligent router to toggle between low-latency chat and deep retrieval; benchmarked at 13 QPS with a minimal 4.2% privacy overhead.
  • Tech: Python, CyborgDB, LangChain, Flask, Vector Encryption.

2️⃣ EviScore | Multilingual Clinical Extraction

  • The Build: Architected a pipeline to extract structured data from multilingual (Hinglish) medical journals using Sentence-Transformers.
  • Key Innovation: Eliminated hallucinations by engineering Pydantic guardrails and a Semantic N-gram Scorer (mBERT) to ensure cross-lingual term grounding.
  • Tech: Python, Sentence-Transformer, Pydantic, LangChain, Multilingual-BERT.

3️⃣ Celeb Look-Alike | Metric Learning & CV

  • The Build: Fine-tuned ResNet50 and implemented MediaPipe for face alignment, standardizing inputs against pose and lighting variations.
  • Key Innovation: Boosted accuracy from 45% to 83.4% through layer unfreezing and data augmentation; utilized L2-Normalized Embeddings for precise one-shot matching.
  • Tech: Python, TensorFlow, MediaPipe, Scikit-Learn, Flask.

🌱 Currently Learning

  • Advance RAG
  • Agentic AI

📫 Connect with me

LinkedIn(#)

Pinned Loading

  1. FinGuard FinGuard Public

    Python 1

  2. EviScore EviScore Public

    Python

  3. celeb_look_alike celeb_look_alike Public

    It genrates the best match of input face to a celebrity

    Python 1

  4. Legal_Assistance Legal_Assistance Public

    Its a RAG application build as a assistance/chat-bot especially for legal purposes/query.

    Jupyter Notebook