ML Engineer · Computer Vision, ASR · ML Systems & Inference Optimization · Robotic Perception
"Building ML systems that ship — from training to optimizing low-level CUDA kernels."
ML Engineer completing M.S. in CS at University at Buffalo (GPA 3.96), with 6+ years of Research Experience, 3+ years shipping production ML systems. My work lives at the intersection of Model Finetuning, GPU kernel optimization, and efficient model deployment.
- 🎤 CV, ASR & AI — UNet, Yolo, DeepLab, Wav2Vec, Conformer, Diffusion Models, LLMs, HuggingFace, Pytorch, TensorFlow, Nemo
- ⚡ GPU Systems — CUDA kernels, TensorRT, QAT/PTQ, AWQ/GPTQ/SmoothQuant, NVCC, PySpark, GCP Dataproc
- 🔬 Research — 7 peer-reviewed publications across IEEE IGARSS, BMVC, SDSC; 20+ citations
- 🚀 Open to Applied ML Engineer · ML Systems Engineer · Research Engineer . Perception roles from July 2026 (STEM OPT)
| Project | Description | Status |
|---|---|---|
| AutoPhon | Phoneme-level speech annotation tool using Wav2Vec 2.0 + Whisper; reduces annotation time 10× | 🟢 Active |
| BiDAAM | Post-hoc interpretability framework for diffusion models | 📄 Accepted at IEEE SMC 2026 |
| TF-Net | Multi-task satellite image segmentation; SOTA 94% F1 on SpaceNet2 + WHU | ✅ IEEE IGARSS 2025 |
Computer Vision and 3D
Speech & ASR
GPU & Systems
Quantization & Compression
ML Frameworks
Languages
| # | Paper | Venue | Year |
|---|---|---|---|
| C.1 | BiDAAM: A Bi-directional Attribution Framework for Diffusion Models | IEEE SMC | 2026 |
| C.2 | A Tuning-Fork Network for Improved Building Footprint Extraction | IEEE IGARSS | 2025 |
| C.3 | Evaluating Cooling Efficacy of Urban Green Spaces During Extreme Heat Events | IEEE IGARSS | 2025 |
| C.4 | Unsupervised Landmark Discovery Using Consistency Guided Bottleneck | BMVC | 2023 |
| C.5 | Improved Flood Mapping for Efficient Policy Design | IEEE IGARSS | 2023 |
| C.6 | PD-SEG: Population Disaggregation Using Deep Segmentation Networks | IEEE IGARSS | 2023 |
| C.7 | Estimating Spatio-Temporal Urban Development using AI | SDSC | 2022 |
University at Buffalo (SUNY) Aug 2024 – Jul 2026
Graduate Research Assistant
· ASR → Reduced PER 80%→15% on children's speech via Wav2Vec 2.0 + AutoPhon pipeline
· GPU → CUDA kernels for transformer attention/FFN on A100; profiled vs cuBLAS
· Quant → Outlier-aware QAT for Conformer ASR: 7.17 WER vs 93.7 naive INT8
· Bench → AWQ/GPTQ/SmoothQuant across A100, Jetson, Raspberry Pi
CITY at LUMS Aug 2021 – Jul 2024
ML Research Engineer
· CV → Satellite Image Segmentation, Vehicle Detection and Classification, Route Optimization
· Deploy → TensorRT on NVIDIA AGX Xavier: 5× throughput (3→14 FPS); Punjab Safe Cities Authority
· Publish → TFNet: A novel architecture for Building Footprint Extraction; 94% F1 on SpaceNet2; IEEE IGARSS 2025
· Prod → City-scale waste routing platform: ~100K liters/month fuel savings (LWMC)
Information Technology University Sep 2020 – Jul 2021
Graduate Research Assistant
· 3D → DSAC* for camera pose estimation; Colmap for 3D Reconnstruction
· CV → Unsupervised facial landmark extraction; GNN, GCN
Available for full-time roles from July 2026 · STEM OPT · Buffalo, NY → Open to relocation