ARAVIND PAI
Bengaluru
Summary
Generative AI Research & Development Technical Lead at Caterpillar with the proven track record of building Large Language Model (LLM) based products. Expert in finetuning Large and Small Language Models, Agentic Systems, Retrieval Augmented Generation (RAG) and Prompt Engineering. Demonstrated consistent success in scaling the AI prototypes to production-grade solutions.
Overview
6
6
years of professional experience
Work History
SENIOR DATA SCIENTIST
Caterpillar
Bengaluru
04.2024 - Current
- Led the development of two flagship Generative AI products at Caterpillar - Cat AI Assistant and In Cab Assistant
- Developed Cat AI Assistant with 95% trajectory accuracy, unified digital interface enabling customers and dealers to interact seamlessly with various applications of Caterpillar, utilising Langgraph orchestration framework and multi-agentic architecture to address user queries effectively.
- Built In Cab Assistant, a real-time voice-based assistant deployed on NVIDIA Jetson Thor hardware, leveraging finetuned Small Language Models (SLMs) for complex tool calling and facilitating multi-turn conversations through vLLM achieving 50 milliseconds latency.
- Spearheaded development of RAG-based solution for troubleshooting guides, employing document summary retriever to achieve 95% context and answer relevancy.
SENIOR DATA SCIENTIST
Analytics Vidhya
Gurugram
05.2021 - 03.2024
- Spearheaded the development of AI products, including DHSGPT and high-performance Recommendation System.
- 2x improvement in lead generation with DHSGPT, a Retrieval Augmented Generation assistant utilising an auto-merging retriever that achieved 98% accuracy in both context and answer relevancy metrics.
- Enhanced pages per session by 50% on blog platform using semantic search-based recommendation system with embedding model to suggest top k articles and evaluated using Recall@k.
DATA SCIENTIST
Spektacom
Bengaluru
07.2020 - 04.2021
- Developed end-to-end computer vision minimum viable product to detect front foot no balls in cricket, analysing live video streams from cameras positioned beside popping crease for real-time detection, deployed on RTX 3070 hardware.
Education
MSC - Data Science
PSG College of Technology
Coimbatore01-2021
Skills
- Finetuning Large and Small Language Models
- Agentic AI
- Retrieval Augmented Generation (RAG)
- Natural Language Processing (NLP)
- Machine Learning
- Deep Learning
- Generative AI
- Prompt Engineering
- Python
- Unsloth Framework
- Transformers Reinforcement Learning(TRL)
- Langgraph
- Autogen
Timeline
SENIOR DATA SCIENTIST
Caterpillar
04.2024 - Current
SENIOR DATA SCIENTIST
Analytics Vidhya
05.2021 - 03.2024
DATA SCIENTIST
Spektacom
07.2020 - 04.2021
MSC - Data Science
PSG College of Technology