Abinaya Mahendiran

I’m currently the CTO of Nunnari Labs (Coimbatore) where I set up and manage the technological strategy, orchestrate the development of cutting-edge AI and ML solutions while staying attuned to the ever-evolving landscape of the AI/ML domain. That involves crafting the roadmap for product innovation, fostering a culture of research-driven development, and nurturing a high-caliber technical team. Alongside it, I consult independently on end-to-end NLP, ML/DL and MLOps work.

I’m a hands-on engineer based out of India with 13 years of experience across software engineering and data science, focused on building enterprise-ready as well as technology-for-good AI solutions. My areas of expertise include NLP/NLU/NLG, Deep Learning, Machine Learning, Explainability, Generative AI and MLOps/LLMOps, and I have worked both as an individual contributor and as a manager building products and teams — previously at Mphasis NEXT Labs, Cisco and Cognizant. I love contributing to open source and have published several research papers as a result.

From 2023 to 2024 I was the part-time Program Manager at IITM for an open source project called Hidden Voices, created to reduce the gender data gap on Wikipedia.

I also co-lead an open source community called AI Tamil Nadu for the past six years and have been contributing back to the community through mentoring students and young professionals, by writing technical articles and organizing events and sharing knowledge through technical talks. I take up paid part-time consultation via topmate.

latest posts

Aug 28, 2023	Hidden Voices: Biographies of women achievers that are hidden no more!
Feb 02, 2023	Machine Learning Model Training Mistakes: How to avoid them
Jan 19, 2023	Avoiding Machine Learning Pitfalls: From a practitioner’s perspective — Part 5

selected publications

NL-Augmenter: A Framework for Task-Sensitive Natural Language Augmentation

Kaustubh D. Dhole, Varun Gangal, Sebastian Gehrmann, and 123 more authors

2022

Abs

Data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Python-based natural language augmentation framework which supports the creation of both transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of natural language tasks. We demonstrate the efficacy of NL-Augmenter by using several of its transformations to analyze the robustness of popular natural language models. The infrastructure, datacards and robustness analysis results are available publicly on the NL-Augmenter repository
GEMv2: Multilingual NLG Benchmarking in a Single Line of Code

Sebastian Gehrmann, Abhik Bhattacharjee, Abinaya Mahendiran, and 74 more authors

2022

Abs

Evaluation in machine learning is usually informed by past choices, for example which datasets or metrics to use. This standardization enables the comparison on equal footing using leaderboards, but the evaluation choices become sub-optimal as better alternatives arise. This problem is especially pertinent in natural language generation which requires ever-improving suites of datasets, metrics, and human evaluation to make definitive claims. To make following best model evaluation practices easier, we introduce GEMv2. The new version of the Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers to benefit from each others work. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.
Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning

Shivalika Singh, Freddie Vargus, Daniel Dsouza, and 30 more authors

2024

Abs

Datasets are foundational to many breakthroughs in modern artificial intelligence. Many recent achievements in the space of natural language processing (NLP) can be attributed to the finetuning of pre-trained models on a diverse set of tasks that enables a large language model (LLM) to respond to instructions. Instruction fine-tuning (IFT) requires specifically constructed and annotated datasets. However, existing datasets are almost all in the English language. In this work, our primary goal is to bridge the language gap by building a human-curated instruction-following dataset spanning 65 languages. We worked with fluent speakers of languages from around the world to collect natural instances of instructions and completions. Furthermore, we create the most extensive multilingual collection to date, comprising 513 million instances through templating and translating existing datasets across 114 languages. In total, we contribute four key resources: we develop and open-source the Aya Annotation Platform, the Aya Dataset, the Aya Collection, and the Aya Evaluation Suite. The Aya initiative also serves as a valuable case study in participatory research, involving collaborators from 119 countries. We see this as a valuable framework for future research collaborations that aim to bridge gaps in resources.