Ishaan Watts

AI Researcher

Pittsburgh, Pennsylvania, United States2 yrs 1 mo experience

Key Highlights

Expert in Machine Learning and NLP techniques.
Contributed to significant AI research publications.
Developed innovative solutions for multilingual challenges.

Stackforce AI infers this person is a Machine Learning expert with a focus on AI research and applications in various domains.

Contact

wattsishaan18@gmail.com LinkedIn

Skills

Core Skills

Machine LearningNatural Language Processing (nlp)Data Science

Other Skills

Anomaly DetectionArchitectural ImprovementsAzure DatabricksBenchmarkingCollaborationContinual LearningData AnalyticsData EngineeringData MiningDataset CreationDeep LearningEvaluationsFair Evaluation PlatformsFraud DetectionGraph Networks

About

Hello! My name is Ishaan Watts and I am a first-year graduate student in the Machine Learning department at Carnegie Mellon University. I am interested in building self-improving systems which can evolve over time and in exploring reinforcement learning techniques to reason efficiently. In a past life, I was a Pre-Doctoral Researcher at Google DeepMind under the guidance of Dr. Partha Talukdar. I was part of the 'Modular Large Scale Continual Learning' group, where I explored model composition. I also worked as a Research Intern at Microsoft Research under the guidance of Dr. Sunayana Sitaram, where we tackled the linguistic diversity challenges in India and worked towards fairness in evaluations [NAACL, ACL, EMNLP, AAAI]. I completed my bachelor's at IIT Delhi. Check out my Personal Website (https://wattsishaan.github.io) for more details. Outside of research, I’m a bit of a fitness freak and I enjoy going to the gym or do distance-running in my free time. I also love watching cricket! I am currently on the lookout for ML internships in the US for Summer 2026. Please feel free to contact me if there are any relevant opportunities! Last Updated: 23/08/2025

Experience

Google deepmind

Pre Doctoral Researcher

Jul 2024 – Jul 2025 · 1 yr · Bengaluru, Karnataka, India · On-site

Part of the 'NLP team' under the guidance of Dr. Partha Talukdar and the 'Modular Large Scale Continual Learning' group headed by Dr. Marc'aurelio Ranzato.
Explored model composition through architectural improvements, extensive evaluations, and explorations into multi-model stitching.
Contributed to Gemini 2.5 Pro (https://arxiv.org/abs/2507.06261v1).

Model CompositionArchitectural ImprovementsEvaluationsMulti-Model StitchingMachine LearningNatural Language Processing (NLP)

Microsoft

NLP Research Intern

May 2023 – Jun 2024 · 1 yr 1 mo · Bengaluru, Karnataka, India · On-site

Guide: Dr. Sunayana Sitaram
1. PARIKSHA 📚: We built a fair and transparent evaluation platform for 10 Indic languages, collaborating with People+ai and Karya. Preprint - https://lnkd.in/gUSnN7RK
2. MAPLE 🔍 : We explored the QLoRA finetuning technique and evaluated over 90 models on multilingual datasets. Accepted at ACL Findings 2024 - https://lnkd.in/gyWsA62h
3. MEGAVERSE 📊 : We benchmarked the multilingual capabilities of over 20 LLMs across 23 datasets and 83 languages, and did a contamination study. Accepted at NAACL 2024 - https://lnkd.in/gXwCg6V2.
Collaborated with Dr. Adrian de Wynter from Microsoft Redmond on RTP-LX 🤬, where we released a culturally nuanced toxic multilingual dataset in 28 languages. Preprint - https://lnkd.in/gTrPNY5T
I also worked with Dr. Akshay Nambi and Dr. Tanuja Ganu on Shiksha Copilot 🤖. We created a web app to help teachers design engaging content and deployed it in schools across Karnataka.

Fair Evaluation PlatformsMultilingual DatasetsToxic Multilingual DatasetWeb App DevelopmentNatural Language Processing (NLP)Machine Learning

Torch investment management

Machine Learning Engineer

Sep 2022 – Dec 2022 · 3 mos · Noida, Uttar Pradesh, India · Remote

Mentor: Mr. Amit Sharma
Project Title: Stock Price Modelling 📈
Refactored LightGBM model codebase to predict Top30 US S&P500 stocks and modelled Saudi market data.
Developed a new feature using NLP techniques to determine the correlation between stock price and tweet sentiment.
Scraped Twitter using snscrape, and performed topic-based filtering using BART and sentiment analysis using FinBERT to create the feature.

LightGBMNLP TechniquesTwitter ScrapingSentiment AnalysisMachine LearningData Science

Udaan.com

Data Scientist

May 2022 – Jul 2022 · 2 mos · Bengaluru, Karnataka, India · Remote

Mentor: Mr. Pranjal Singh
Project Title: Holistic User-Embeddings via GNNs
Developed framework to generate holistic user-embeddings from buyer-seller interaction graph for better segmentation.
Built complex multi-relational & multi-entity graph and modeled Hetero-Graph AutoEncoder with a novel loss function.
Improved Udaan fraud detection by 2.45% using generated embeddings in the deployed PAFv2 model.

User-EmbeddingsGraph Neural NetworksFraud DetectionData ScienceMachine Learning

Griffith university

Research Intern

May 2021 – Jul 2021 · 2 mos · Queensland, Australia · Remote

Guide: Dr. Saiful Islam
Project Title: Malware Detection using Deep Learning
Performed malware detection and program analysis of binaries from VirusShare using deep learning techniques.
Constructed Control Flow Graphs from malware binaries through static analysis and used opcodes as features for nodes.
Applied tf-idf vectorisation on dataset & designed Graph Convolutional Network to achieve 89.1% accuracy.

Malware DetectionDeep LearningStatic AnalysisMachine LearningData Science