██╗ ██╗██╗██╗ ██╗███████╗██╗ ██╗ ███████╗██╗ ██╗ █████╗ ██████╗ ███╗ ███╗ █████╗
██║ ██║██║██║ ██║██╔════╝██║ ██╔╝ ██╔════╝██║ ██║██╔══██╗██╔══██╗████╗ ████║██╔══██╗
██║ ██║██║██║ ██║█████╗ █████╔╝ ███████╗███████║███████║██████╔╝██╔████╔██║███████║
╚██╗ ██╔╝██║╚██╗ ██╔╝██╔══╝ ██╔═██╗ ╚════██║██╔══██║██╔══██║██╔══██╗██║╚██╔╝██║██╔══██║
╚████╔╝ ██║ ╚████╔╝ ███████╗██║ ██╗ ███████║██║ ██║██║ ██║██║ ██║██║ ╚═╝ ██║██║ ██║
╚═══╝ ╚═╝ ╚═══╝ ╚══════╝╚═╝ ╚═╝ ╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝
██╗ ██╗██╗██╗ ██╗███████╗██╗ ██╗
██║ ██║██║██║ ██║██╔════╝██║ ██╔╝
██║ ██║██║██║ ██║█████╗ █████╔╝
╚██╗ ██╔╝██║╚██╗ ██╔╝██╔══╝ ██╔═██╗
╚████╔╝ ██║ ╚████╔╝ ███████╗██║ ██╗
╚═══╝ ╚═╝ ╚═══╝ ╚══════╝╚═╝ ╚═╝
███████╗██╗ ██╗ █████╗ ██████╗ ███╗ ███╗ █████╗
██╔════╝██║ ██║██╔══██╗██╔══██╗████╗ ████║██╔══██╗
███████╗███████║███████║██████╔╝██╔████╔██║███████║
╚════██║██╔══██║██╔══██║██╔══██╗██║╚██╔╝██║██╔══██║
███████║██║ ██║██║ ██║██║ ██║██║ ╚═╝ ██║██║ ██║
╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝╚═╝ ╚═╝
VIVEK SHARMA
VIVEK SHARMA
vivek@fedora:~$ whoami
> Jr. Software Engineer - Machine Learning @ Anvex AI Technologies
> Built RAG systems, LLM applications, and custom NLP models
> BSc IT Graduate | Linux User | Open-Source Contributor
vivek@fedora:~$ ls -la
/home/vivek/contacts/
vivek@fedora:~$ wget
/home/vivek/documents/resume.pdf
vivek@fedora:~$ cat
/etc/passwd | grep vivek
I'm Vivek Sharma, a Jr. Software Engineer specializing in Machine Learning at Anvex AI Technologies. With a BSc in Information Technology from the University of Mumbai, I build production-ready RAG systems, develop custom LLM applications, and architect NLP models from scratch. My work spans telephony AI platforms, policy chatbots, and custom sentence embedding models. As a passionate Linux user and open-source contributor, I leverage PyTorch, LangChain, and vector databases to create intelligent systems that solve real-world problems.
vivek@fedora:~$ ls -la
~/projects/
MedVec-Scratch/
pytorch
transformers
bpe-tokenizer
Custom sentence embedding model with Siamese Transformer architecture (4 layers, 4 heads, 256 dim).
Features: Custom BPE tokenizer (30k vocab), trained on 233k medical triplets, Triplet Margin Loss
git clone
colab notebook
ArtCycle/
python
tensorflow
cyclegan
Artistic style transfer between photos and paintings using unsupervised deep learning.
Status: deployed on Streamlit Cloud
git clone
FederalRegistryRAG/
groq-llm
mysql
streamlit
RAG-based AI system to query US Federal Registry documents.
Features: real-time data, async processing, clean UI
git clone
StudentAnalytics-SPAPS/
xgboost
mongodb
streamlit
ML-powered academic performance prediction system.
Features: interactive dashboards, secure role-based access
git clone
NeuroScan/
cnn
tensorflow
huggingface
MRI classification system for detecting brain tumors using pre-trained CNN.
Status: live on Hugging Face Spaces
git clone
MobileUserBehavior/
ann
tensorflow
streamlit
Streamlit app predicting user behavior using mobile usage data.
Features: trained ANN model, visual progress indicators
git clone
OnlineSales-EDA.ipynb
pandas
matplotlib
seaborn
Exploratory analysis and data preprocessing on online sales dataset.
Output: insights, visualizations, engineered features for modeling
view notebook
vivek@fedora:~$ cat
/var/log/experience.log
Anvex AI Technologies Private Limited - Jr. Software Engineer Machine Learning
September 2025 – Present | Mumbai, India
Project: AnvexSpeak
- Built RAG system for AnvexSpeak, a telephony AI platform using Twilio/Plivo with multi-source data extraction
- Implemented BAAI/bge-large-en-v1.5 embeddings with advanced filtering for client databases, Google Drive, and web scraping
- Created custom LLM wrapper supporting Gemma3, Qwen, Mistral, Llama, and OpenAI using LangChain/LangGraph
- Deployed with Qdrant (production) and ChromaDB (development) for real-time document-grounded responses
Anvex AI Technologies Private Limited - Machine Learning Intern
June 2025 – August 2025 | Mumbai, India
Project: AVA
- Developed AVA, a RAG-powered policy chatbot using LangChain/LangGraph for query routing
- Integrated Qdrant vector store and SambaNova Llama-4 across 26+ policies
- Built FastAPI endpoints achieving 800-1500ms response time with confidence scoring
- Implemented advanced query routing and document retrieval systems
Techathon - Mulund College of Commerce
February 2025 | Mumbai, India
- Built predictive maintenance models using sensor data and historical maintenance records
- Performed feature engineering and hyperparameter tuning for model optimization
- Integrated ML models into backend using FastAPI for real-time predictions
- Achieved accurate equipment failure predictions for preventive maintenance
vivek@fedora:~$ ps aux |
grep skills
Frameworks
- tensorflow-gpu
- pytorch
- flask
- fastapi
- langchain
- langgraph
Libraries
- scikit-learn
- numpy
- pandas
- matplotlib
- seaborn
- plotly
- streamlit
DevOps
- docker
- git
- vim
- bash
- jupyter-lab
Fedora Linux 42 | Kernel 6.15.7-100.fc41.x86_64 | Shell: bash 5.2.32
Uptime: 1 hour, 29 mins | Memory: 5.36 GiB / 7.40 GiB (73%)
vivek@fedora:~$ echo "© 2025 Vivek Sharma - Built with ❤️ and lots of ☕"
Connection to vivek02sharma.github.io established. Welcome to my digital
workspace!