Ashu Gulati
,

Results-driven AI Researcher & Developer with 4 years of hands-on experience. Seeking to leverage my skills and experience in advance AI (Artificial Intelligence) technologies using Large Language Models (LLMs) and Natural Language Processing (NLP).

Experience


Lead Software Engineer (AI Research)
November 2020
 — 
Present
Persistent Systems

Contributed on multiple projects in AI research and development in the domain of business process automation using AI.

  • Developed BERT-based intent classification models using RASA agents for conversational AI interface in a multi-turn dialogue system.
  • Curated and evaluated a travel data use case to survey the usefulness of autonomous AI agents to generate a sequence of actions in enterprise setting. Reported improvements in bleu and F1 score.
  • Created an end-to-end pipeline from dataset collection to evaluation of the LLM (large language model) generated results using Python scripts and customized metrics. This helped in identifying the performance of differently sized and trained LLMs.
  • Trained teacher-student model using PyTorch with a sentiment classification task, applying curriculum learning in the dataset to distill large-sized model performance in small-sized model.
  • Benchmarked a dataset for evaluation on function calling tasks used in enterprise setting.

Projects


Customer Support, Tune2Precision Hackathon
July 2024
 — 
August 2024

Fine-tuned an open-source LLM for tailored customer support in the travel industry, enhancing response accuracy, and relevance.

Policy Recommendation, GenAI Hackathon
August 2023
 — 
November 2023

Integrated a traditional insurance policy recommendation system with a large language model (LLM) to improve explainability.

Trassist, IBM Call for Code
August 2023
 — 
October 2023

Developed a mobile application for waste collection and segregation leveraging IBM Services and Hugging Face model.

Context-Retaining Chatbot
August 2023
 — 
September 2023

Built a context-aware conversational chatbot for question answering using multiple documents and Retrieval-Augmented Generation (RAG) techniques.

SynthAI, Semicolons Hackathon
March 2023
 — 
February 2023

Led a team of 8 in creating a framework for generating synthetic, domain-agnostic datasets using GPT3.

Education


SPPU (Sinhgad Institutes, Pune) - 8.56 GPA
July 2016
 — 
August 2020
Bachelors of Engineering (B.E.) in Information Technology
Delhi Public School, Ranchi - 90.1%
July 2014
 — 
April 2016
12th in Science

Certificates


LLM102x: Large Language Models: Foundation Models from the Ground Up
May 2024
edX
LLM101x: Large Language Models: Application through Production
March 2024
edX

Publications


Co-authored and presented at EMNLP 2023
December 2023
ACL (Association for Computational Linguistics)

Explored challenges and research opportunities in using LLM-based autonomous agents for tool composition in enterprise settings.

Awards


Bravo - Team Award
September 2024
Persistent Systems

A total of four team awards and two individual awards received appreciating my ability to learn and adapt quickly.

2nd Runner-Up, Tune2Precision Hackathon
August 2024
Persistent Systems
Runner-Up, GenAI Dash Hackathon
November 2023
Persistent Systems
Top Talent Award for FY 2023
June 2023
Persistent Systems

Skills


Programming Languages and Frameworks:
Python, Bash, PyTorch, Transformers, Hugging Face, LangChain
AI Techniques:
Artificial Intelligence (AI), Machine Learning (ML), Natural Language Processing (NLP), Deep Learning, Knowledge Distillation, Transfer Learning, Fine-Tuning, LLM Prompting, Data Processing, Data Analysis
Development Tools:
Jupyter Notebook, VSCode, PyCharm, Git, Docker, UnslothAI, Azure