Shaheen Nabi
Research @ Carnegie Mellon University | LLM Post-Training, RLVR, Mechanistic Interpretability
Pittsburgh, Pennsylvania, United States · Remote
💬About Me
- I am a Researcher at Carnegie Mellon University (remote), working on LLM post-training, reinforcement learning, reasoning, distillation, agentic RL, and RL environments — with research problems centered on improving the reasoning and agentic capabilities of large language models through scalable training and evaluation methods.
- My research spans the full stack around reasoning models — post-training methods that build reasoning (RLVR, preference optimization, reasoning distillation), the test-time strategies that make it efficient (compute allocation, deciding when to stop deliberating), and the empirical/evaluation work needed to trust any of it. Right now I'm going deeper into mechanistic interpretability — studying the internal circuits and representations behind multi-step reasoning in large language models, to understand failure modes like reward hacking and unfaithful chain-of-thought well enough to actually fix them, not just detect them. A specific thread I'm pulling on: disentangling memorization from genuine reasoning mechanistically — which circuits are doing lookup versus computation — and using that understanding to work toward guiding reasoning safely, so models can be steered away from unfaithful or unsafe reasoning traces rather than just penalized after the fact.
- I build and release everything in the open — training/eval systems, checkpoints, and datasets — in lieu of formal publications, via GitHub and Substack.
📝Publications
- Beyond Fixed Directions: Adaptive Representation Analysis of Reasoning and Memorization in LLMs An empirical study of whether reasoning and memorization can be characterized by a single representation direction and whether that geometry remains stable after GRPO.
- (in-coming)
💻Experiences
- 2026.08 - now: Researcher, LLM Post-Training & Reasoning, Carnegie Mellon University, Pittsburgh, PA (Remote).
- 2026.06 - 2026.08: Researcher, Mechanistic Interpretability & Reasoning, Indian Institute of Science (IISc), Bengaluru.
- 2025.12 - 2026.05: Member of Technical Staff, RL Environments & Post-Training for Coding Agents, Stealth AI Startup, Singapore (Remote).
- 2025.05 - 2025.11: Machine Learning Engineer, Agentic Systems, Stealth AI Startup (Travel-Tech), Virginia, USA (Remote).
- 2024.11 - 2025.04: Machine Learning Engineer, Agentic Systems, Stealth AI Startup (Travel-Tech), Ontario, Canada (Remote).
- 2024.05 - 2024.10: Machine Learning Intern, Computer Vision, iNeuron.ai, Bengaluru.
- 2023.01 - 2023.12: Machine Learning Engineer, NIELIT, Jammu & Kashmir (Hybrid).
- 2021.01 - 2022.12: Software Engineer, Lasso Pacific.
- 2019.07 - 2020.12: Software Engineer (Intern → Full-time), Shexion, Jammu & Kashmir.
🛠️Selected Projects
- Open Post-Training A single reusable system that takes a base model through the full post-training recipe, stage by stage, rather than one-off scripts per experiment: evaluation, preference tuning, reasoning distillation, and RLVR training all share the same components. I use it as a research testbed — running hypotheses through the pipeline and producing demos/proof-of-output for specific ideas. One such run: a GRPO ablation on a base model (64.0% MATH-500) dropped to 48.0% at a 5-step checkpoint before recovering to 60.0% at 50 steps, with response length shrinking from 303 → ~220 tokens — evidence that sparse binary-correctness reward is a weak signal without more steps and better reward shaping.
- Olmo-3 From Scratch From-scratch implementation of AllenAI's Olmo-3 (7B/32B) — GQA, YaRN-scaled RoPE (8K→65K), sliding-window attention with periodic global refresh, SwiGLU, RMSNorm — plus a custom KV-cache module with unrotated key/value storage for RoPE compatibility, reducing cache memory via GQA while supporting bfloat16 inference.
- Reinforcement Learning: Zero to Hero A maintained RL repository spanning MDPs through PPO and DDPG, with clean implementations and math annotations for practitioners moving into post-training research.
- LeafLogic Dataset 25K annotated crop/plant images, open-sourced on Hugging Face — downloaded by thousands of developers around the world.
📖Education
- Bachelor of Computer Applications — Information Technology, Machine Learning, Deep Learning & AI Systems