Skip to content

Swethanjali Brahma

Generative AI & ML Engineer Intern • Srikakulam District, Andhra Pradesh, 532445 • s***********@gmail.com • +91*******769 • drivetube.ai/•••••

Professional Summary

Generative AI Engineer with 1+ years of experience building LLM-based RAG pipelines, local LLM chatbots, and supervised ML systems. Experienced with LangChain, LlamaIndex, HuggingFace embeddings, Chroma vector DB, LangSmith observability, and end-to-end ML lifecycle from data ingestion and feature engineering to model evaluation and deployment.

Technical Skills

Programming Languages: Python,Java
Frameworks and Libraries: Pandas,NumPy,Scikit-Learn,TensorFlow,Keras,LangChain,Matplotlib,Seaborn
Databases: SQL,HuggingFace Embeddings,Chroma,Vector Search,Semantic Search
Cloud and DevOps: AWS
Data and Analytics: Unstructured,CharacterTextSplitter,XGBoost,Feature Engineering,Model Evaluation,Cross-Validation,Hyperparameter Tuning
Tools and Methodologies: LlamaIndex,Ollama,OpenAI API,Agentic AI,Prompt Engineering,RAG Architecture,LangSmith,Git,GitHub
Deep Learning and NLP: Transformers,SpaCy,CNN,RNN
Speech and Interaction: SpeechRecognition,pyttsx3,Text-to-Speech

Work Experience

Naresh IT
Hyderabad, India
Generative AI & Agentic AI Intern (Data Scientist / ML Engineer Intern)
Ongoing
Worked at an IT training and software services organization building Generative AI, RAG, and ML solutions for document intelligence and student support applications.
Tech Stack: Python, LangChain, LlamaIndex, Groq Llama 3.3-70B-Versatile, HuggingFace Embeddings, Chroma, LangSmith, Ollama, OpenAI API, Unstructured, CharacterTextSplitter, SpeechRecognition, pyttsx3, Scikit-Learn, XGBoost, TensorFlow, Pandas, NumPy, Git, AWS
  • Built an end-to-end Retrieval-Augmented Generation (RAG) pipeline using LangChain and LlamaIndex with Groq Llama 3.3-70B to extract, chunk, and summarize information from complex PDF documents for document intelligence workflows.
  • Architected vector ingestion and preprocessing pipelines using Unstructured data loaders and CharacterTextSplitter, storing embeddings in Chroma to enable low-latency semantic search across large document sets.
  • Implemented HuggingFace embeddings for semantic similarity and integrated Chroma vector DB to support precise context retrieval that improved relevant context matching for downstream prompts.
  • Designed and integrated LLM observability using LangSmith to trace chain execution, evaluate prompt chains, and monitor latency, enabling iterative performance tuning and failure analysis.
  • Engineered a localized Ollama-based chatbot for real-time student psychological support, integrating speech-to-text and text-to-speech with SpeechRecognition and pyttsx3 to provide responsive voice interactions.
  • Developed and validated supervised ML classifiers for mental-health signal detection; evaluated Logistic Regression, SVM, Decision Trees, KNN, and XGBoost, achieving 95% classification accuracy and robust cross-validated performance.

Projects

LLM-Powered Document Intelligence & RAG Pipeline
Tools Used: LangChain, LlamaIndex, Groq Llama 3.3-70B-Versatile, Unstructured, CharacterTextSplitter, HuggingFace Embeddings, Chroma
  • Implemented end-to-end pipeline to ingest, chunk, embed, and semantically search PDF documents to enable LLM summarization and question answering over enterprise documents.
  • Optimized preprocessing to minimize contextual loss and tuned embedding strategy to improve retrieval relevance for RAG prompts.
Ollama AI Student Support Chatbot
Tools Used: Ollama, LangChain, SpeechRecognition, pyttsx3, Prompt Engineering
  • Engineered a low-latency, fully localized chatbot to provide psychological and stress support to students, combining text and voice interfaces for accessibility.
  • Improved contextual query accuracy to 93% by implementing strict NLP preprocessing and advanced LLM reasoning prompts.
Detection of Psychological Instability Using Machine Learning Algorithms
Tools Used: Scikit-Learn, XGBoost, Feature Engineering, Cross-Validation, Pandas, NumPy
  • Developed statistical classification engines to identify mental-health risk indicators using high-dimensional behavioral datasets, achieving 95% classification accuracy.
  • Streamlined ML lifecycle with feature scaling, cross-validation, and model selection across Logistic Regression, SVM, Decision Trees, KNN, and XGBoost to prevent overfitting.

Education

Sri Sivani College of Engineering (JNTUGV)
Bachelor of Technology, Computer Science and Engineering • Andhra Pradesh, India • 2024
Thamminaidu Junior College
Intermediate Education • Andhra Pradesh, India • 2020
Thamminaidu Concept School
Secondary School Certificate (SSC) • Andhra Pradesh, India • 2018

Certifications

Data Science Trainee Certification — Quality Thoughts, Hyderabad, India • Aug 2025 - May 2026

Powered by Drivetube · Create your own profile at drivetube.ai