Harshdeep Gupta

Software Engineer

Seattle, Washington, United States7 yrs 6 mos experience

AI ML PractitionerHighly Stable

Key Highlights

7+ years of experience in building distributed systems
Led successful migration to Azure Kubernetes Service
Developed machine learning applications for resume enhancement

Stackforce AI infers this person is a Cloud Infrastructure and Machine Learning specialist with a focus on distributed systems.

Contact

Skills

Core Skills

Distributed SystemsC++KubernetesAzureMachine LearningData VisualizationWeb Development

Other Skills

PythonTypeScriptARM templatesPyTorchTableauSQLTensorFlowNLPNode.jsHTMLCSSBashGNU DebuggerGrafanaAzure Kusto

About

I have 7+ years of experience building early-stage products and shipping them from zero to one. As an early member of a distributed systems team at Microsoft, I helped ship a successful v0 product and have worked on infrastructure products at Azure scale. My technical expertise includes modern C++, full-stack development, and distributed systems.

Experience

7 yrs 6 mos

Total Experience

7 yrs 6 mos

Average Tenure

7 yrs 6 mos

Current Experience

Microsoft

2 roles

Software Engineer II

Sep 2021 – Present · 4 yrs 9 mos · On-site

As an early engineer on a 'zero-to-one' distributed compute platform, I helped build a new, next-generation compute orchestration layer for Microsoft. My work involved deep dives into container internals and virtualization primitives to build highly efficient, core systems in modern C++.

PythonDistributed SystemsC++

Software Engineer

Dec 2018 – Sep 2021 · 2 yrs 9 mos · On-site

During my tenure at Azure Monitor, my team maintained the critical infrastructure backbone for services that ingest petabytes of data daily.
My most significant initiative was leading the migration of our cluster infrastructure to Azure Kubernetes Service (AKS). This project was a major win, resulting in a combined 33% cost reduction and 10% performance increase while serving 100% of production traffic. I specifically engineered the AKS clusters using ARM templates, which enabled efficient, automated rollouts across 30+ production regions.
My focus on stability also led to a critical refactoring of the Cache Invalidation Service, which drove a 100% reduction in recurrent incidents.

KubernetesTypeScriptAzure

National university of singapore

Deep Learning for Course Recommendation (NUS Internship)

Jul 2018 – Oct 2018 · 3 mos · Singapore · On-site

I designed and built a job-course recommender to help Singapore residents discover job-ready training. I wrote a Neural Collaborative Filtering model in PyTorch, tackling an extremely sparse interaction matrix (~99.99%), working with original paper authors. Before modeling, I ran exploratory data analysis (Python/SQL) to understand user/course distributions, cold-start pockets, and signal quality. To boost relevance, I added metadata embeddings (age, gender, Area of Training) and a multi-task objective that jointly learned relevance and category, with regularization to counter sparsity.
For communication and decision-making, I created Tableau dashboards to visualize model behavior, adoption paths, and lift over baselines, then presented the findings to government officials at SkillsFuture Singapore. The final system delivered an ~40% error reduction versus an item-popularity baseline and provided clear next steps for integrating metadata and evaluation into the program’s rollout.

PyTorchTableauMachine LearningData Visualization

Vmock

Summer Intern

May 2017 – Jul 2017 · 2 mos · Gurgaon, India

The project was to build a suite of machine learning based applications to improve quality of english in resumes. We did this using deep learning for NLP in TensorFlow.
The first project was to build a Grammar Corrector System. We trained LSTMs using seq2seq models on resume data, which was augmented by introducing artificial errors. For generating the training data, we extended a rule based grammar corrector using custom rules for resume English. The rule based system was developed and improved to become a beta version in itself.
The second project was to build a spelling corrector system. This was achieved by taking a conventional spelling corrector system and augmenting it with RNN to suggest contextually most probable replacement.

Microsoft

Software Developer Internship

May 2016 – Jul 2016 · 2 mos · Hyderabad, India

As a part of visual studio team, built a VS Code extension to enhance publishing experience on marketplace. The extension generates live preview of details page(the page visible to a normal user wanting to install a extension), thus replacing a three step loop with a continuous feedback system.
Development of this extension involved writing asynchronous code in Typescript and working largely with NodeJS, HTML, CSS and Git. Also, the code was tested extensively with unit tests written using Mocha, Chai and Sinon. The build chain was automated using Travis CI.
The project is live on VS marketplace, with more than 3k downloads and multiple positive reviews.