cv

Basics

Name Zoher Kachwala
Label PhD Candidate in LLM Post-Training & Evaluation
Url https://zoher15.github.io/
Summary PhD candidate specializing in LLM post-training and evaluation. Designed REMATCH (NAACL'24), a novel AMR graph evaluation metric achieving 5x speedup while ranking first in semantic similarity. Developed Prefill-Guided Thinking (NeurIPS'25 Workshop), achieving 24% F1 improvement for zero-shot AI image detection. Build and deploy systems using PyTorch, vLLM, and multi-GPU infrastructure.

Education

Publications

Work

  • 2022.08 - 2024.12
    Teaching Assistant
    Introduction to Network Science
    Led lab sections and designed exam problems and homework assignments covering network fundamentals including graph traversal, community detection, shortest paths, and hub analysis.
  • 2020.08 - 2020.12
    Teaching Assistant
    Applied Machine Learning
    Supported course on classical machine learning methods including linear and logistic regression, decision trees, and ensemble methods.
  • 2019.08 - 2021.12
    Teaching Assistant
    Elements of Artificial Intelligence
    Designed homework assignments and built autograding infrastructure to support 300+ graduate students on foundational AI topics including search algorithms, Markov models, heuristics, and decision trees.

Skills

Research
LLM Post-Training
Model Evaluation
Benchmarking
Supervised Fine-Tuning (SFT)
Prompt Engineering
Chain-of-Thought Reasoning
Multimodal AI
RLHF
Model Interpretability
Programming
Python
PyTorch
C++
CUDA
JavaScript
SQL
Bash
ML Tools
PyTorch
Hugging Face Transformers
vLLM
Weights & Biases (W&B)
TensorFlow
Scikit-learn
Infrastructure
Multi-GPU Training
GPU Clusters
Distributed Computing
AWS
Google Cloud Platform (GCP)
Batch Processing
Data/Eval
Benchmark Development
Statistical Analysis
Data Pipeline Design
A/B Testing
Pandas
NumPy