Pranav Das

Pranav Das

B.Tech Computer Science graduate
India
English, Malayalam
Contract On-site

About Me

B.Tech Computer Science graduate with hands-on experience building end-to-end ML pipelines and LLM-powered applications. Built RAG systems using LangChain, ChromaDB, Ollama, and Google Gemini - including a local document…

Experience

Vice-Chairperson → Chairperson

IEEE Computer Society SB CEC · Chengannur , Kerala , India
Jan 2024 - Jan 2026 · 2 years

<p>Grew active CS membership from 73 to 398 (445% growth).</p><p> Organized 15 events (up from 5 in 2023), including 7 women-focused events centered on empowerment and skill-building, in collaboration with IEEE Computer Society Kerala Section and partner colleges.</p><p> Secured a $900 grant from IEEE Computer Society.</p><p> Revamped the DevCom Community into a dedicated platform for development and collaboration.</p><p> Elected Chairperson in February 2025 after this track record.</p><p> Led executive, design, and technical teams to plan and deliver workshops and events.</p><p> Mentored junior volunteers.</p><p> Drove cross-team collaboration across the chapter.</p><p> Secured Best Student Branch Chapter Award and Best Event Award (ARCANE Hackathon) from IEEE Computer Society Kerala Section.</p>

PROJECTS

Customer Churn Prediction System

NA · https://customer-churn-prediction-system-ml-project.streamlit.app/
Duration : 20-Jun-2026 - 26-Jun-2026

<p>Built an end-to-end telecom churn prediction system using Python, XGBoost, Scikit-learn, and</p><p>Streamlit. Handled class imbalance via SMOTE, applied label encoding and data preprocessing on</p><p>the Telco dataset, and tuned hyperparameters using RandomizedSearchCV - achieving a weighted F1</p><p>score of 0.91 on a heavily imbalanced dataset. Deployed as an interactive Streamlit web app supporting</p><p>real-time predictions and customer risk assessment.</p>

PDF QnA Bot

NA · https://pdf-qna-bot.streamlit.app/
Duration : 13-Jun-2026 - 19-Jun-2026

<p>Built an end-to-end RAG-based PDF Question Answering chatbot using Python, LangChain, Google</p><p>Gemini, and Streamlit. Implemented a full retrieval-augmented generation pipeline - loading and</p><p>chunking PDFs via PyPDFLoader and RecursiveCharacterTextSplitter, embedding chunks using Gemini</p><p>Embeddings, storing vectors in an InMemoryVectorStore, and retrieving top-k semantically similar</p><p>chunks at query time. Integrated Gemini 2.5 Flash as the LLM with a context-grounded prompt to prevent hallucination. Deployed as an interactive Streamlit web app with persistent chat history and</p><p>real-time document Q&A.</p>

Local RAG AI Agent

NA · https://github.com/Pranavdas-1/local-rag-agent
Duration : 06-Jun-2026 - 12-Jun-2026

<p>Built a local Retrieval-Augmented Generation (RAG) system that answers natural language questions</p><p>about restaurant reviews using LangChain, ChromaDB, and Llama 3.2 via Ollama. Converted CSV</p><p>review data into vector embeddings using mxbai-embed-large, stored in a persistent ChromaDB vector</p><p>database. Implemented semantic retrieval to fetch top-5 relevant reviews per query, passed as context</p><p>to a locally running Llama 3.2 model to generate grounded, hallucination-guarded answers.</p>

Domain-Specific Chatbot Fine-Tuning (Qwen2.5)

NA · https://github.com/Pranavdas-1/vedaz-astrologer-finetune
Duration : 31-May-2026 - 05-Jun-2026

<p>Fine-tuned Qwen2.5-1.5B-Instruct using LoRA on a 55-conversation Hindi/Hinglish astrology chat</p><p>dataset, training under 0.3% of total parameters for efficient domain adaptation Built a data-cleaning</p><p>pipeline to parse malformed JSON conversation logs into HuggingFace Dataset format for supervised</p><p>fine-tuning via TRL’s SFTTrainer Debugged multi-GPU device conflicts, CUDA compatibility issues,</p><p>and dependency errors across Kaggle/Colab environments to get training running end-to-end</p>

Certifications

NA

na · Aba , Sichuan , China · 2026

Skills

Python JavaScript SQL Java Machine Learning Regression Classification Feature Engineering Data Preprocessing EDA Hyperparameter Tuning RandomizedSearchCV Cross Validation Model Evaluation RAG Vector Embeddings Prompt Engineering LoRA Fine-Tuning Supervised Fine-Tuning Pandas NumPy Scikit-learn XGBoost LangChain ChromaDB Streamlit Matplotlib Seaborn SciPy PyTorch HuggingFace Transformers PEFT TRL Git GitHub Ollama Jupyter Notebook Gemini API Kaggle Google Colab
Report this Profile?