A little about me.A few things
I’ve made.

(vibed by Codex )

I’m Pakaphol Thadawasin,
usually go by Big.

Born and raised in Bangkok, Thailand. I spent my college years in Cambridge, MA, earning a BS and MEng at MIT. Now based in Bellevue, WA, I’m a Machine Learning Applied Scientist at Microsoft.

My focus

Large-scale LLM
post-training.

From data and reward design to infrastructure and training recipes.

↓
01 /

About & experience

Big sitting with a sleeping cat on his lap

Education

Massachusetts Institute of Technology

Cambridge, MA

MEng in Computer Science and Engineering

Aug 2024–May 2025

Selected coursework: Theoretical Machine Learning, Deep Learning, TinyML, and Advanced Complexity Theory.

BS in Computer Science and Engineering (6-3)

Aug 2021–May 2024

Selected coursework: Algorithms, Machine Learning, Natural Language Processing, and Theory of Computation.

Experience

  1. Apr 2026–present

    Microsoft

    Machine Learning Applied Scientist

    I currently focus on large-scale LLM post-training (RLHF/RLVR) and knowledge distillation for Copilot in Microsoft 365 apps, with an emphasis on data, reward construction, and reinforcement learning environments. Past work includes agent evaluation and benchmarking.

  2. Jul 2025–Apr 2026

    TikTok

    Machine Learning Applied Scientist

    In customer service, issues can quickly affect customer satisfaction, so detecting and understanding them in production is essential. I developed features that connected different service domains within TikTok’s seller customer-service chatbot, giving users a unified experience. I also worked on online evaluation, using production experiments to guide decisions and quickly detect and analyze issues affecting the customer experience.

  3. Sep 2024–May 2025

    MIT Edelman Lab

    Research Assistant

    I researched how to capture the interplay between genotype and phenotype, advised by Prof. Elazer Edelman. One approach involved training models and using explainable AI methods to visualize and interpret the relationships they learned. This included developing evaluation and gene-attribution methods for single-cell foundation models. I contributed to this research under the mentorship of Farhan Khodaee and Rohola Zandie, who later co-founded Absentia Labs, building on this broader research direction.

02 /

Selected work

Questions explored.
Ideas built.

01EmailBenchPaper · arXiv preprint · 2026

A Benchmark for Evaluating LLM Agents on Enterprise Email and Productivity Tasks. We introduce 206 scenarios across 16 task categories to evaluate whether email agents complete tasks, combining executable checks with language-model grading.

Read paper
02Gradient-Based Gene SelectionPaper · ICLR 2025 MLGenX Workshop · Tiny Paper Track

Gradient-Based Gene Selection for scRNA-seq Foundation Models. Research on gradient-based attribution for selecting informative genes and interpreting what a single-cell foundation model learns.

Read paper
03The Dissipation Theory of AgingPaper · npj Aging · 2025

My contribution focused on quantitative analysis, including entropy analysis and plots showing how entropy distributions evolve with age across tissues and cell types.

Read paper
04LLM Quantization & Inference EfficiencyProject · MIT 6.5940 TinyML

I explored 4-bit weights and activations for LLaMA through post-training quantization and quantization-aware training. I also wrote ARM NEON and x86 AVX2 kernels for INT4 matrix multiplication in on-device LLaMA2-7B inference, using vectorized dequantization and multithreading.

05How Well Have VAEs Learned Latent Space?Project · MIT 6.7960 Deep Learning

I investigated disentanglement in the latent representations learned by variational autoencoders (VAEs), examining how they represent underlying factors of variation.

03 /

Outside of work

Yellowtail kimbap at Jungsik
Yellowtail KimbapJungsik, New York
Caviar and cornbread at Birdsong
Caviar and CornbreadBirdsong, San Francisco

Always up for
a good meal.

In my spare time, I enjoy trying restaurants, from local spots to Michelin-starred fine dining. Food is also a way for me to learn about different cultures. Each cuisine carries its heritage, reflecting the local environment and how people live.

There’s a lot to discover through food. And, of course, it’s delicious.

Follow along with my experiences on Instagram.

@thecuratedbites

A few highlights

Michelin-starred dining

And the list goes on…

04 /

Get in touch

Want to Connect?

  • Personal
  • Work
  • Elsewhere
  • Research