Leo Vaychik
home>work>contractor
Back to profile

Sembly AI

Backend & MLOps Engineer

SA
PeriodJul 2021 – Aug 2023
LocationRemote
ImpactMLflow + DVC pipeline modularization
Websitesembly.ai/
Tech Stack
PythonDjangoDVCMLflowMS Teams APIPostgreSQLDockerMLOps

About the Company

Sembly AI is an AI-powered meeting assistant platform that transcribes, summarizes, and extracts action items from corporate voice and video conferences.

My Journey There

I started at Sembly AI on the Backend team, shipping Django user stories, maintaining internal dashboards, error tracking, and testing tools.

My first major initiative was researching third-party APIs and building an automated agent capable of connecting to live MS Teams sessions. I integrated it into the core backend and ran the first live client demo.

Later, I transitioned to the MLOps team to tackle machine learning infrastructure bottlenecks. I introduced DVC for tracking dataset and model weight versions, and used MLflow to break our monolithic ML pipeline into clean microservices — making model delivery and inference serving across staging and production significantly faster.

Key Deliverables & Architectures

  • [01]MS Teams Bot & Core API Integration: Researched third-party APIs and built an automated agent capable of joining live MS Teams conference sessions; delivered the first client demo and implemented Django core API endpoints.
  • [02]MLOps Microservices (MLflow): Transitioned to the MLOps team and decoupled a monolithic machine learning pipeline into modular microservices using MLflow, significantly accelerating model deployment and inference across environments.
  • [03]Data & Model Versioning (DVC): Implemented DVC (Data Version Control) to track dataset iterations, model artifacts, and pipeline dependencies, establishing reproducible ML experiment workflows.
  • [04]Backend Hygiene & Internal Tooling: Implemented Django user stories, maintained internal testing tools, error-tracking pipelines, and operational dashboards for backend services.