Shaheen Nabi
Hi! 👋 I'm Shaheen, an independent Research Engineer based in Bengaluru, India. My research is centered on reasoning and thinking models — understanding what makes a model reason well, and how to make it reason better at lower cost.
I approach this from the ground up: studying how decisions made during pre-training and mid-training — data composition, continual pre-training, and architectural choices like hybrid MoE and state space models — shape the reasoning capabilities that emerge downstream. I'm interested in how these foundation-level choices propagate through to post-training, and what they make possible or limit.
The core of my work lives in post-training: RLVR, SFT, preference optimization pipelines, and test-time compute — how to get a model to think better, for less. I actively follow and engage with new research in reasoning: small reasoning architectures, hierarchical reasoning, and emerging approaches to efficient thinking. The longer-term question I care about is how these directions can push the next generation of frontier reasoning models — cracking harder benchmarks while becoming more compute-efficient at inference.
I contribute to open source through code and writing, and am actively working toward pushing contributions into large LLM infrastructure and post-training frameworks.
Research Interests
- Scaling Foundation Models Data composition, quality, and curation at scale. How pre-training decisions — compute allocation, data mixtures, and training dynamics — shape the capabilities available to post-training.
- Continual & Mid-Training How continued pre-training and mid-training interventions affect the reasoning qualities that emerge in later stages. The relationship between training phase, data distribution, and downstream reasoning behavior.
- Architectures for Reasoning How architectural choices — hybrid MoE, state space models, and dense-sparse combinations — affect reasoning capabilities. Studying new open model families to understand what architectural decisions make models more or less amenable to post-training reasoning improvements.
- Post-Training Pipelines RLVR, SFT, and preference optimization. How reward signals, data quality, and training dynamics in post-training interact with the capabilities laid down during pre-training.
- Test-Time Compute & Efficient Reasoning Making models think better at lower cost. Process reward models, adaptive compute budgeting, and search-guided generation. The core question: how do we build models that know when to stop thinking, not just how to think longer.
- Emerging Reasoning Architectures Actively following new directions: small reasoning models, hierarchical reasoning architectures, and novel thinking paradigms. Interested in how these approaches can inform the next generation of frontier reasoning models and open new possibilities on hard benchmarks.
Outputs
Open-source repositories, writing, and models — in lieu of publications.