Ikteder Akhand Udoy

PhD student in Computer Vision and Data Science at Boise State (LPiNS Lab)

I build reliable ML pipelines and evaluation tooling clean data ingestion, reproducible experiments, and deployment aware verification. Lately I’ve also been working on LLM workflows and evaluation for my Po2-QAT research.

Computer Vision GenAI / LLMs Model Efficiency (Po2-QAT) Data Pipelines Edge / GPU

Education

Current degree and background.

Boise State University

PhD in Computing (Computer Vision & Data Science) · 2023–Present

LPiNS Lab · Advisor: Dr. Omiya Hassan

American International University Bangladesh

BSc in Computer Science & Engineering · 2017–2021

Strong CS fundamentals and software projects

Experience

Research, teaching support, and industry experience.

Graduate Research Assistant · Boise State University (LPiNS Lab)

2023–Present · Boise, Idaho

Research + engineering work in data pipelines, deep learning experiments, reproducibility, and deployment-aware validation. TA/teaching support is part of the same role.

Python Pipelines Deep Learning Reproducibility LLM Workflows

Built Python pipelines to ingest and standardize multi-source datasets into a consistent schema.
Designed ablations and logged metrics (CSV/JSON) for clear A/B comparisons and reporting.
Implemented export + verification scripts to replay runs and catch mismatches before deployment.
Supported algorithms/data structures/digital systems coursework: debugging, code reviews, Linux lab support.
Delivered an IEEE TinyML workshop (2025): data collection → training → on-device testing with repeatable handouts.

Junior Project Manager · Playense

2021–2023 · Remote

Coordinated software tasks and kept delivery organized through clear tracking and communication.

Coordination Documentation Delivery

Tracked issues, clarified requirements, and communicated status to technical/non-technical stakeholders.
Helped teams stay aligned on priorities and timelines through concise reporting.

Projects

Research pipelines, tools, and hands-on builds.

Shift-Add Inference Simulator (Po2-QAT)

Exports compact model packages and replays layer outputs to validate correctness before deployment.

VerificationQuantizationPython

Vision Training Stack

Reproducible classification/segmentation with ablations, clean logs, and export-friendly packaging.

PyTorchCVExperiments

Dataset QA Toolkit

Schema checks, missingness, split integrity, and distribution sanity checks to prevent silent data bugs.

Data QApandasReliability

Multi-Dataset Medical Vision Pipeline

Unified preprocessing/metrics across CheXpert, ChestMNIST, BreastMNIST, EchoNet-Dynamic, and more.

Medical AIPipelinesPython

LLM Evaluation + RAG Prototypes

Task-driven evaluation, retrieval baselines, and careful logging for repeatable comparisons.

LLMsRAGEval

TinyML Workshops

Keyword spotting, gesture, and object detection: guided data collection → training → on-device testing.

TinyMLEdgeTeaching

Skills

Things I use most often.

Core

Languages: Python, Java, C++, SQL, Bash
CS: data structures, algorithms, debugging, clean code
Workflow: Git/GitHub, reproducible experiments, documentation

ML / Systems

Deep learning: PyTorch, TorchVision, TensorFlow (familiar)
Data: pandas, numpy, matplotlib, CSV/JSON logs
Platforms: Linux, Colab/Jupyter, Docker (basics)