Yash Jain
Welcome! I am a Research Scientist at Essential AI Labs, where I build foundation models alongside Ashish Vaswani (Attention is All You Need). I was a core contributor to Rnj-1, an open-source coding and agentic foundation model with over one million downloads across Hugging Face and Ollama — I worked on execution-based code generation, the agentic tool-calling stack, and the long-context mid-training that grew into Rnj-1.5 (32k→160k tokens, 40% on SWE-bench Verified).
Previously, I was an ML Scientist II at Microsoft, where I worked on diffusion models and multimodal large-language models in collaboration with Vibhav Vineet and Harkirat Behl, publishing at NeurIPS, CVPR, ICLR, ICCV, NAACL, EMNLP, ISWC, and UbiComp.
I graduated from Georgia Tech with an M.S. in Computer Science, advised by Zsolt Kira. Before that, I earned my B.Tech. in Computer Science from IIT Bombay, where I received an Excellence in Research Award under Soumen Chakrabarti.
Reach out by email if you wish to collaborate!
news
| Jul 22, 2026 | Featured in The Times of India (in print, across India) — an article on my journey and the research behind frontier AI labs! (thread) |
|---|---|
| May 25, 2026 | Released Rnj-1.5, a long-context extension of Rnj-1 scaling to 160k tokens! By moving from all-global to block-local/global attention layers, our 8B model hits 40% on SWE-bench Verified (thread). |
| Dec 09, 2025 | Released Rnj-1, a state-of-the-art open-source coding and agentic foundation model with 600k+ downloads on Hugging Face! |
| Jun 16, 2025 | Joined Essential AI Labs as a Research Scientist! |
| Mar 13, 2025 | Local Prompt Optimization Paper accepted at NAACL 2025 for Oral Presentation (Main Conference)! |
| Jun 05, 2023 | Joined Microsoft as an ML Scientist II at Redmond! |
selected publications
-
Multi-Stage Multi-Modal Pre-Training for Automatic Speech RecognitionIn Joint International Conference on Computational Linguistics and Language Resources and Evaluation (LREC-COLING), 2024
service
- Conference reviewer: NeurIPS (2024, 2025, incl. Datasets & Benchmarks track), CVPR (2025; AI for Content Creation workshop, 2024), ICCV (2025), ECCV (2024), ICML (2024, 2025), ICLR (2025, 2026), ACL Rolling Review (2025–2026)
- Program Committee: AAAI 2026
- Journal reviewer: Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2023–2026