# Sumanth R Hegde > LLMs @ Anyscale || Prev: C3 AI, UC San Diego, IIT Madras Location: San Francisco, California, United States Profile: https://flows.cv/sumanthrhegde I'm currently a software engineer at Anyscale, working in the LLM team. Broadly, I'm passionate about machine learning. I'm currently building SkyRL: https://github.com/NovaSky-AI/SkyRL Previously, I did my master's in computer science at UCSD. I've worked on a bunch of projects in natural language processing, computer vision and computer systems in my undergraduate and graduate studies. I've also made open-source contributions to HuggingFace's PEFT (parameter-efficient fine tuning) and Accelerate libraries. In Summer 2023, I was a Data Science intern at C3 AI, working on fine-tuning language models for their Generative Search application. I also have previous industry experience as a machine learning intern at Hakimo and HyperVerge. Much of my work in undergrad was in computer vision, and I also have a publication in UDC Workshops, ECCV 2020. ## Work Experience ### Software Engineer @ Anyscale Jan 2024 – Present | San Francisco, California, United States Software Engineer working on the LLM team at Anyscale! Current Project: SkyRL; Skythought - One of the core contributors of SkyRL: https://github.com/NovaSky-AI/SkyRL , building a full stack library for post-training LLMs - Core contributor to https://novasky-ai.notion.site/skyrl-v0 - implemented a scalable remote server for RL training on SWE-Gym, contributed to building asynchronous multi-turn rollout implementation to improve SWE-Bench performance of Qwen-3-8B by 5.8%. - Core contributor to novasky-ai.notion.site/skyrl-sql, one of the first open-source models trained with multi-turn RL on Text2SQL - matching GPT-4o and o4-mini on the Spider benchmark. - One of the core maintainers for the Skythought repo: https://github.com/NovaSky-AI/SkyThought/commits?author=SumanthRH - Worked on standardized, scalable evaluation for reasoning models Past Project: LLMForge, Anyscale's fine-tuning framework - Added support for different fine-tuning tasks (such as instruction tuning and causal LM) - Improved model source support to allow bringing any HuggingFace model with any chat template to fine-tune on Anyscale. - Preference tuning, Function calling fine-tuning - Improved DPO training speed by 20-40% with prefix sharing: https://github.com/frankxwang/dpo-prefix-sharing - Led building an SDK for models trained on Anyscale: https://docs.anyscale.com/reference/llm_models ### Student Researcher @ UC San Diego Jan 2023 – Jan 2024 Working on language models in Prof. McAuley's Lab at UCSD ### Graduate Teaching Assistant @ UC San Diego Jan 2022 – Jan 2024 - Teaching Assistant for CSE 232: Principles of Database Systems and CSE 21: Mathematics for Algorithms and Systems. - Responsibilities included conducting weekly discussion sessions, preparing question papers for examinations, etc. Best part: Office hours! ### Data Science Intern @ C3 AI Jan 2023 – Jan 2023 | Redwood City, California, United States - Set up a finetuning codebase for language models from scratch for use in C3’s Generative Search application - Features: Support for difference causal and sequence-2-sequence models, ability to mix different training datasets (for a text-to-text or a causal language modelling task), visualize metrics on multiple evaluation datasets, parameter-efficient fine-tuning and quantization support, etc - Trained 10B+ parameter models on 1M+ samples using DeepSpeed and 🤗 Accelerate. ### Machine Learning Engineer Intern @ Hakimo Jan 2023 – Jan 2023 Worked on video-based object detection methods to improve Hakimo's Remote Guarding solution ### Undergraduate Student Researcher @ Indian Institute of Technology, Madras Jan 2020 – Jan 2021 – Demonstrated fast reconstruction of a 12 frame video from a single image of a lensless camera, reducing inference time from 2 hours to 30 milliseconds. – Proposed an efficient reconstruction framework - a physics-aware neural net trained in an adversarial fashion, used feature-based loss for producing photorealistic videos. My Bachelor's thesis can be found here: https://tinyurl.com/sumanth-btech-thesis ### Undergraduate Student Researcher @ Indian Institute of Technology, Madras Jan 2020 – Jan 2020 - Created a novel deep learning based model for image restoration, resulting in a publication at ECCV Workshops 2020 and placed 2nd /150 teams at the Under Display Camera Challenge. – Developed a two stage pipeline for directly processing megapixel images with a simulation scheme for data augmentation. – Rectified severe blur and low light conditions in the images, obtaining >12% improvement in image quality with 88% (7.8M) lesser parameters than existing work. ### Deep learning intern @ HyperVerge Inc. Jan 2019 – Jan 2019 | Bengaluru Area, India - Implemented a learning-based face detection algorithm for Know-Your-Customer services, reduced false positives 10 times and false negatives by 2.5 times on HyperVerge’s benchmark. – Trained a Multi-task Cascaded Convolutional Neural Network using > 200,000 images to beat the previous model which had >99.5% accuracy. – Analysed client data and employed hard positive mining, data augmentation to improve recall by 5% . ## Education ### Master's degree in Computer Science UC San Diego ### Bachelor of Technology - BTech (Honours) in Electrical Engineering Indian Institute of Technology, Madras ## Contact & Social - LinkedIn: https://linkedin.com/in/sumanthrhegde - Website: https://github.com/sumanthrh --- Source: https://flows.cv/sumanthrhegde JSON Resume: https://flows.cv/sumanthrhegde/resume.json Last updated: 2026-04-05