ML Engineer · Buffalo, NY

Muhammad
Ahmad
Waseem.

I build production ML systems — speech recognition, CUDA kernels, and quantized models that run fast on real hardware. Currently finishing my M.S. in Computer Science at University at Buffalo, where I research ASR for children's speech and efficient inference.

7 peer-reviewed publications. 3+ years shipping models from research to government-scale deployment. Available July 2026.

7
Publications
3.94
GPA · M.S.
TensorRT speedup
§ 01 / SELECTED WORK

Things I've built
that matter.

A cross-section of six projects spanning speech recognition, GPU systems, computer vision, and production geospatial tooling. Each ships real metrics — not demos.

All projects →
§ 02 / NOW

What I'm working on.

Live snapshot of current research threads at UB. Updated weekly.

This week
Outlier-aware QAT for Conformer ASR — pushing INT8 below 7.5 WER without custom kernels.
Active
This month
Writing the reverseDAAM camera-ready for IEEE SMC 2026.
Review
This semester
Profiling custom attention kernels against cuBLAS and FlashAttention-2 on A100.
Active
On deck
Porting AutoPhon inference to Jetson Orin for on-device phoneme feedback.
Planning
§ 03 / REFERENCES

What advisors say.

Three years of close research collaboration, in their words.

You bring a lot of experience to our team and your presence allows other team members to grow as well.
Dr. Momin Uppal
Associate Professor · LUMS
Your exceptional skills in computer vision give us a lot of confidence in picking new projects.
Dr. Muhammad Tahir
Associate Professor · LUMS
You have been a very hard-working student and your never-give-up attitude is not found in many people.
Dr. Arif Mahmood
Thesis Supervisor · ITU