Dreycey Albin

Dreycey Albin

Senior ML Software Engineer
at Microsoft Azure

About Me

ML engineer with roots in computational biology, focused on bridging research and production systems. I design and validate traditional machine learning and deep learning models, build experimentation frameworks, and ship low-latency inference pipelines. At Microsoft Azure, I work at the intersection of modeling and distributed systems on Resource Central, the central ML platform in the control plane, serving all regions and handling more than 1 million requests per day.

New York City, NY

education
Ph.D. Computer Science
University of Colorado Boulder · NSF GRFP Fellow
2020 – 2023
M.Sc. Systems, Synthetic & Physical Biology
Rice University
2018 – 2020
B.S. Chemistry + B.S. Biology
University of Northern Colorado · McNair Scholar
2012 – 2017
interests
Computational BiologyGenerative AIProbabilistic Machine LearningML SystemsDistributed Systems

Work Experience

Microsoft Azure

New York, NY
Senior Machine Learning Engineer, Level 63 08/2026 – Present

Led a cross-functional effort to move Azure from uniform to per-VM oversubscription by predicting VM CPU utilization at creation time, a cold-start problem with no usage history. Built a LightGBM quantile regression model with a tunable risk-capacity tradeoff, maintaining ≤1% under-prediction and a 10 ms inference SLO in online performance tracking. Integrated predictions across systems into the Azure scheduler and business logic, owning the full lifecycle from modeling and production inference to capacity-saving decisions.

Machine Learning Engineer II, Level 62 03/2025 – 08/2026

Cut regional capacity mitigation time from ~1 week to ~4 hours across 50+ regions and ~1M VMs/containers, with zero customer impact, by building a telemetry-driven mitigation system; earned Azure Core Impact and Builders Excellence awards. Architected a distributed framework for automated A/B and shadow evaluation of candidate ML models across Resource Central; patent filed.

Machine Learning Engineer II, Level 61 05/2023 – 03/2025

Drove platform-wide changes that improved Azure server-packing efficiency through intelligent, safe overpacking of internal workloads. The project required forecasting fleet-wide impact from historical signals to balance expected savings against operational risk, followed by staged production validation and pipelines to measure realized savings for senior leadership.

Medtronic

Boulder, CO
Research Software Engineer, Contract 09/2021 – 05/2022

Developed a real-time LSTM pose estimator from fiber-optic sensor streams for surgical catheter tracking, and shipped the full production Python software stack for an autonomous catheter robot including real-time control, telemetry, and failure-safe behaviors that passed regulatory-readiness review.

Software / Projects

EnrichSeq

EnrichSeq

A bioinformatics pipeline for phage enrichment analysis.

Python / Nextflow / Bash
PhageBox

PhageBox

Embedded system for bacteriophage research automation.

C/C++
PhageScanner

PhageScanner

A reconfigurable machine learning pipeline for labeling ORFs/proteins in bacteriophage genomes and metagenomic data.

Python
PhageFilter

PhageFilter

PhageFilter uses a Sequence Bloom Tree (SBT) to filter bacteriophage reads from metagenomic files.

Rust
Metscale

Metscale

MetScale performs metagenome comparison estimates, taxonomic classifications, and functional predictions on metagenomic data

Snakemake / Python
SeqScreen

SeqScreen

SeqScreen sensitively assigns taxonomic classifications and functional annotations to short and long (ont) DNA sequences.

Python / Bash / Nextflow
Kvar

Kvar

A pipeline for finding disease-associated kmers in genomic data

Bash / Python / R

Contact

Feel free to reach out for collaborations or questions.