Some of the work below was developed in private enterprise repositories and later upstreamed to public repositories by my team. Public commit history may not reflect individual authorship.
Repo: https://agenttoolkit.github.io/agent-lifecycle-toolkit
Tech: Python, Langchain
What I worked on:
- Designed and implemented ToolGuard component for enforcing LLM/Agent adherance to user policies
- Work described in ArXiV preprint: https://arxiv.org/abs/2510.14842, under conference review
Repo: https://github.com/ibm/m3-train-eval
Tech: Python, Deepspeed
What I worked on:
- Designed and ran multi-GPU tuning experiments (LoRA/ALoRA) for models with transformer, MoE, and State-Space architectures
- Included DPO-style training capability
- Created tool-calling agent and evaluation harness to validate training efficacy
Repo: https://github.com/IBM/UQ360
Tech: Python, Uncertainty Quantification
What I worked on:
- Led development of black-box meta-model uncertainty quantification algorithm
- Lead author on AAAI accepted publication describing this approach: https://arxiv.org/abs/2012.08625
Repo: https://github.com/pmservice/NeuNetS
Tech: Python, Jupyter
What I worked on:
- Contributed as member of dev team delivering component as production feature