Hey there! I'm fascinated by how machines can understand human language, and that's what I do for a living. I build systems that can read, understand, and answer questions just like humans do (well, almost!). Think of it as teaching computers to be better conversation partners.
While I geek out on Machine Learning and AI in general, what really gets me excited is working with different languages. Telugu, my mother tongue, holds a special place in my heart. I'm on a mission to make AI work better for languages that big tech often overlooks. It's like bringing the power of AI to the languages that deserve more attention!
When I'm not diving deep into code or playing with language models, you'll find me hitting the pavement for a long run or strumming away on my string instruments. Running helps me clear my head, and music... well, let's just say my instruments are patient with my amateur attempts at making music! 😄
Systems and methods for intelligent query processing using natural language understanding and knowledge graphs. Tokenizes user queries, identifies intents, and executes relevant functions to generate contextual responses using enterprise data sources.
Converts natural language instructions into executable software transactions. Uses NLP to transform user inputs into application-specific formats, filters based on context parameters, and matches to appropriate transactions for execution.
Safety and control mechanisms for large language model deployments in enterprise environments.
A suite of LLM-based modeling copilots for text-to-model translation, spanning zero-shot, chain-of-thought, and agentic strategies, with an online leaderboard for benchmarking execution and solution accuracy.
@misc{text2model,
title = {Text2Model: Modeling Copilots for Text-to-Model Translation},
author = {Serdar Kadıoğlu and Karthik Uppuluri and Akash Singirikonda},
year = {2026},
eprint = {2604.12955},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2604.12955}
}
Fine-tuning open-weight small language models, from 0.6B to 20B parameters, to generate syntactically valid MiniZinc constraint models directly from natural language text.
@misc{learn2zinc,
title = {Learn2Zinc: Fine-tuning Small Language Models for Text-to-Model Translation in MiniZinc},
author = {Serdar Kadıoğlu and Karthik Uppuluri},
year = {2026},
eprint = {2607.20456},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2607.20456}
}
A novel cross-domain dataset for bridging natural language and constraint optimization modeling using MiniZinc.
View on arXivJournal paper presenting a named entity recognition approach to extract optimization components from natural language problem descriptions.
View on SpringerConference paper introducing the Ner4Opt framework for automated optimization model generation from natural language.
View on SpringerPoster presentation on leveraging natural language processing for optimization problem formulation.
View Poster (PDF)AI Magazine 2025 & IAAI 2025: Open-source AI at scale: Establishing an enterprise AI strategy through modular frameworks
Read Article →CORS 2025 Plenary Talk: Decision Making in the Era of Large-Language Models (Alberta, Canada)
View Slides →ICS 2025: Ner2Zinc
View Slides →AAAI 2025: Text2Zinc
View Slides →Koç University Invited Talk: Ner4Opt
View Details →Manabadi Sadassu • July 2024
Presentation on Telugu corpus database, web applications for Telugu NLP, and demos of language models for the Telugu community.
All Things Open RTP Meetup • May 2023
Exploring the landscape of generative AI technologies, applications, and implications for developers and enterprises.
kuppuluri/telugu_bertu
First-ever BERT Masked Language Model trained on the Telugu language from scratch. A foundational model for Telugu NLP tasks, enabling transfer learning for various downstream applications.
kuppuluri/telugu_bertu_tydiqa
Telugu Question-Answering model fine-tuned on Google's TyDiQA dataset. Enables extractive question answering for Telugu text, advancing comprehension capabilities for low-resource languages.
Small language models fine-tuned to generate MiniZinc constraint models from text, from 0.6B to 20B parameters.
RoBERTa models fine-tuned for named entity recognition in optimization problem descriptions.
For the latest models, demos, and experiments, visit my Hugging Face profiles:
NeurIPS 2025
Neural Information Processing Systems - Premier ML/AI Conference
Named Entity Recognition for Optimization - A framework for extracting optimization components from natural language problem descriptions and generating constraint programming models.
A suite of LLM-based modeling copilots for text-to-model translation, with an online leaderboard and an interactive MiniZinc editor. Accepted at the CPAIOR 2026 Master Class.
Experiments in building language models for Telugu from scratch - exploring low-resource NLP techniques for underrepresented languages.
Text Featurization Library - A flexible and easy-to-use library for text featurization with support for various embedding techniques and transformers. Developed at Fidelity Investments; I am one of several contributors, not a lead maintainer.
