Supervised Machine Learning: Regression and Classification

Unsupervised Learning, Recommenders, Reinforcement Learning

Advanced Learning Algorithms

Machine Learning Specialization

Secondary School Certificate, High School

Senior Secondary School Certificate, PCM

Dual Degree

Data Scientist Intern

Objective: Generate time series data from textual prompts and thoroughly analyze multiple representation learning techniques
• Applied Time2vec and CLAP (Contrastive Language-Audio Pretraining) to evaluate their effectiveness in capturing temporal attributes
• Used t-SNE and PCA to qualitatively evaluate the performance of representation learning techniques on synthetic time series signals
• Employed Pinecone as a vector database to store vectors generated using CLAP model, ensuring efficient retrieval and scalability of data

Research Intern

Building the most reliable agentic analytics platform.

Senior Applied AI Researcher

Objective: Worked on a creative engine tool to automate ad creation, enhancing video generation and customization processes
• Designed workflows using Selenium, optimizing Runway interactions by scripting dynamic tool usage for increased efficiency
• Executed multithreading to process multiple video generation tasks concurrently, significantly reducing execution time
• Integrated CLIP model for image retrieval, enhancing video content creation by matching images with specific textual prompts

*Received a Pre-Placement Offer (PPO)

Consultant - Data Science

Objective: Generation of synthetic on-body gait data to enhance behavioral biometric applications using generative AI models
• Developed a methodology for synthesizing Inertial Measurement Unit data for over 1000+ new individuals using the NUS-IGD dataset
• Enhanced MaskGIT in TimeVQVAE, substantially improving the model's ability to generate highly realistic walking action time series data
• Achieved a Frechet Inception Distance (FID) score of 5.22, demonstrating realism of synthetic time series data for biometric applications

Objective: Implemention of multiple recommender systems for courses offered by Zenteiq Edtech, to optimize course discovery
• Executed a user-search Recommender System utilizing TF-IDF for course discovery by aligning user queries with most relevant courses
• Designed a skill-gap analysis system using one-hot encoding and cosine similarity to recommend courses based on user skill profiles
• Integrated MongoDB and ArangoDB for data integration, ensuring efficient storage, retrieval, and management of user and course data

Developed a high-accuracy authorship verification pipeline using advanced RL alignment to reduce hallucinations.

NLP Research Intern

Objective: Development of a chatbot to provide information tailored for small startups, leveraging RAG for interaction capabilities
• Implemented web scraping techniques using BeautifulSoup to extract and analyze textual data relevant to startups from diverse sources
• Adopted BERT (Bidirectional Encoder Representations from Transformers) for efficient data retrieval and analysis of scraped startup data
• Leveraged Qwen LLM to develop a chatbot using the RAG (Retrieval-Augmented Generation) framework, enhancing user interactions

LLM Engineer Intern

Objective: Development of a web platform similar to Hugging Face for public hosting and access to DeepChem models and datasets
• Integrated interfaces with APIs and MySQL databases using Python and SQLAlchemy, accelerating model deployment for drug discovery
• Created end-to-end automated tests using Cypress, significantly improving software reliability and reducing manual testing efforts

Machine Learning Engineer Intern

Engineered a low-latency, high-reliability enterprise analytics service by architecting optimized agentic workflows and an advanced semantic search framework, achieving 1st position on Spider2 with a 96.7% score.

*Received a Pre-Placement Offer (PPO)

Applied AI Research Intern

Applied AI

Know about Varun Agrawal's work experience, education, bootcamps, certifications, and much more. Varun Agrawal