Skip to content
View vsancnaj's full-sized avatar
πŸ’œ
Focusing
πŸ’œ
Focusing

Block or report vsancnaj

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
vsancnaj/README.md

Typing Animation


🌌 About Me

I’m a Machine Learning Engineer focused on NLP, LLMs, and Transformers(more classical approach).
My work blends engineering with research-driven modeling β€” from hierarchical transformers to LLM-assisted dataset generation, entity extraction, and large-scale text processing .

I love building systems that understand messy real-world text, uncover structure, and power analytics, recommendations, and search.

πŸ”­ Current deep-dive:
Designing a context-aware transformer that fuses sentence-level, section-level, and document-level reasoning for robust employer/entity extraction.


🧠 Core Stack & Skills

β”β”βŠ±β‹† Programming β‹†βŠ°β”β”

β”β”βŠ±β‹† Machine Learning & NLP β‹†βŠ°β”β”

β”β”βŠ±β‹† Cloud & Deployment β‹†βŠ°β”β”

β”β”βŠ±β‹† Data Engineering β‹†βŠ°β”β”

β”β”βŠ±β‹† Tools & Environments β‹†βŠ°β”β”


πŸš€ Featured Projects

πŸ”Ή Hierarchical Job Description Modeling (NLP / Transformers / Research)

A multi-level transformer architecture that integrates:
β€’ Sentence context
β€’ Section-level signals (COMPANY, SKILLS, JOB, etc.)
β€’ Document-wide TF-IDF–guided pruning
Designed for accurate entity extraction + employer detection.
Tech: PyTorch, DistilBERT, attention layers, TF-IDF scoring


πŸ”Ή AI Agent Document Analyzer

Local retrieval agent using
LlamaIndex + LangChain + Ollama models
for PDF question answering and contextual summarization.


πŸ”Ή Sephora Hybrid Recommender System

Combined content-based & collaborative filtering
for skincare recommendations with a simple Flask/Streamlit UI.


πŸ”Ή Vision-Based Anomaly Detector

Built an optical flow + LSTM system to detect heavy-object anomalies
in real-time waste-sorting conveyor footage.


πŸ“Š GitHub Stats


πŸ’Œ Connect With Me

✨ Always open to collaborating on ML/NLP research, transformer architectures, and meaningful AI projects ✨

Profile Views

Pinned Loading

  1. anomaly_detection_thesis anomaly_detection_thesis Public

    This repo enhances waste-to-energy processing using LSTM classifiers to analyze high-speed waste motion from cluttered environments. We use optical flow from WasteAnt datasets to detect anomalies t…

    Jupyter Notebook

  2. Vesta-E-Commerce-Fraud-Detection Vesta-E-Commerce-Fraud-Detection Public

    The objective of this repository is to enhance the accuracy of fraud detection mechanisms by examining transaction patterns and feature correlations.

    Jupyter Notebook 1

  3. AI-Agent-Document-Analyzer AI-Agent-Document-Analyzer Public

    This project is an AI-powered document analysis bot designed to process and extract information from PDF documents.

    Python 1 1

  4. Hybrid-Recommendation-System Hybrid-Recommendation-System Public

    A hybrid recommendation system combining Item-Based Collaborative Filtering and Content-Based Filtering to suggest skincare products based on user preferences, product ingredients, and ratings. Fea…

    Jupyter Notebook 2