
Gregory Golberg
Verified Expert in Engineering
Architect and Back-end Developer
Palo Alto, CA, United States
Toptal member since June 13, 2022
An MIT Computer Science graduate and AWS Certified Solutions Architect, Gregory builds AI systems and the engineering practice around them: MCP servers, retrieval services, multi-agent workflows, and skill infrastructure that lets engineers and agents share what they learn. At Voitta AI, which he co-founded, that work is open source. Behind it are 20 years as a hands-on architect: founding CTO at 2 acquired adtech startups, founding engineer at SnapLogic and Romana, chief architect at Jobox.
Portfolio
Experience
- Relational Databases - 20 years
- Java - 18 years
- SQL - 18 years
- Python - 15 years
- HTTP REST - 15 years
- MySQL - 15 years
- AWS Cloud Architecture - 10 years
- Go - 7 years
Work Experience
Founder and Fractional CTO
Method and Apparatus
- Defined a fractional-CTO offer for small businesses whose operations still run on manual documents: replace the paperwork with systems that run themselves.
- Designed and built document-extraction and voice pipelines on AWS Lambda with LLM back ends and a waterfall across model providers for cost and latency control.
- Packaged technical due diligence and cloud cost reviews as offers: architecture and code-quality assessment, technical-debt inventory, AWS/GCP migration planning, and FinOps.
Co-founder
Voitta AI
- Co-founded an open-source agentic tooling company and designed its core services: a retrieval server for agents, a shared skills catalog, and an OpenAPI-to-MCP bridge.
- Built voitta-rag, a Model Context Protocol server that indexes code and document folders and serves semantic and structural search, plus a per-agent memory store to LLM agents.
- Developed skillz, a cross-host catalog of agent skills and plugins with a schema validator, per-host packaging, and a CI version-bump gate, so a technique learned once installs everywhere.
- Created voitta-yolt, which exposes existing OpenAPI services to LLM agents over MCP, so teams do not have to rebuild their tool surface for agents.
- Ran multi-agent orchestration experiments across Claude and Codex sessions, including cross-session messaging and adversarial review, and published the findings.
Principal Software Engineer
ZoomInfo
- Led an ad-tech platform migration from Amazon ECS to EKS across five production services: Terraform workload modules, Karpenter node pools, HPA and KEDA autoscaling, IRSA, and a cross-region Istio service mesh.
- Ramped production traffic with weighted ALB target-group splits, keeping the cutover zero-downtime and rollback a per-region decision.
- Replatformed observability from CloudWatch to Amazon Managed Prometheus and Grafana: dual-emitted metrics from four services, ported 16+ dashboards, and codified alarms in Terraform.
- Designed a Kong and Amazon Cognito JWT authentication migration for a high-traffic internal API, giving each caller its own identity and listener rule so cutover and rollback stayed independent.
- Rebuilt a Kinesis to ORC to S3 and Athena ingestion path: removed a small-files problem, refactored a monolithic writer into tested collaborators, and moved logging to Protobuf on a dedicated stream.
- Owned correctness on a real-time bidding path: consolidated bot detection, retired two third-party dependencies, and added bid gating, no-bid-reason reporting, and alerting on bid ratios.
- Built a shared agentic-engineering plugin that turns operational debugging into reusable, versioned knowledge, and ran an automated adversarial code review campaign across a large pull-request backlog.
- Drove adoption of agentic coding across the engineering organization, standardizing on Cline and Claude Code, running working sessions, and codifying the review and safety practices that made the tools trustworthy for production work.
Advisor | Consultant
DoubleDoor
- Managed an offshore development team to deliver a website based on Vue and Node.js.
- Consolidated disparate deployment using AWS, which resulted in savings and better latency.
- Migrated Heroku to Fargate and MongoDB from MongoDB Cloud to AWS Fargate.
- Helped to implement the advertising using Google (Ads API, Google Tag Manager, Analytics) and Facebook.
Principal Engineer | Tech Lead
DrumWave
- Led the development of the Offers platform, the centerpiece of the company's data ownership platform.
- Proposed, prototyped, and implemented long-running workflows using the Temporal platform to enable easier development of business logic in a distributed, resilient workflow.
- Acted as a subject-matter expert in API development and testing with Postman and OpenAPI. Created documentation, tutorials, in-house training sessions, blueprints, and initial test suites and tools for the microservice-based teams to interoperate.
- Took on various other responsibilities outside the Offers team's duties, including DevOps.
- Proposed, designed, and developed a tool to convert user-defined queries from UI into safe SQL queries to be run against the data lake. Eventually, it was incorporated into the company's BI offering.
- Created a virtualized data lake and query engine to query heterogeneous data sources (SQL, NoSQL) across multiple clouds (AWS, IBM). I created a DSL for querying, access control mechanisms, and protection against sensitive data/PII sharing.
- Unified the development stack by migrating Java and Python microservices to Node.js, NestJS, and TypeScript.
Advisor | Consultant
Around.Trade
- Designed the cloud infrastructure for the product based on AWS and Vercel and created a CI/CD pipeline for it.
- Added support for CockroachDB in addition to PostgreSQL.
- Integrated Clerk, an OAuth provider, for authentication and authorization.
Back-end Developer
FutureExpress LLC
- Created the initial AWS deployment of the back end using ALB, Fargate, CloudFront, S3, and VPC.
- Built a framework for developing and deploying AWS Lambdas with SAM, CloudFormation, and EventBridge.
- Handled GitHub-action-based deployment for the back end and lambdas.
Lead Platform Architect
SteppeChange
- Worked on proprietary ad tech product involving geocoding.
- Owned the implementation of multiple greenfield projects from specification to deployment on GCP.
- Designed and developed a microservice-based system for managing complex ad flows.
Chief Architect
Jobox.ai
- Implemented best practices to improve velocity, code quality, and stability, e.g., CI/CD, automated unit and integration tests, code coverage, observability, monitoring and alerting, on-call rotation, vulnerability detection, load testing, and more.
- Drove adoption of CI/CD, including automated regression tests and promotion, on-call, and monitoring/observability.
- Led design, development, and delivery of new algorithms for efficient dispatching of home service professionals.
- Cooperated with PMS and EMS to balance internal engineering needs (as above) and product-oriented projects. Developed technology roadmap and AWS cloud infrastructure plan to align with the product roadmap.
Principal Software Engineer
AppLovin
- Joined when AppLovin acquired MZ and also integrated MZ's (the acquired company) existing RTB bidder functionality with AppLovin.
- Worked on green-fielding a non-RTB DSP demand-side platform for partners' use.
- Took over ETL/data pipeline projects and associated tooling.
Staff Software Engineer
MZ (acquired by AppLovin)
- Led maintaining and improving a core marketing and growth engineering functionality, real-time bidder (RTB), accountable for millions of queries per second and millions of dollars spent monthly.
- Operated on the architectural guidance within the product and across teams, including introducing regression testing and ad hoc AWS querying to supplement on-premise clusters for business users and data science.
- Accepted some of the previously external operations roles to become complete product owners, including migration from on-premise to cloud (AWS) and adding DevOps automation.
- Created a lakehouse using Glue and Athena to supplement Hadoop (in AWS) and Vertica (on-premises) for business users and data science.
- Liaised with relevant teams such as release engineering, DevOps, ad operations, and data science.
Software Development Engineer 3
Amazon
- Worked on providing forecasts of available advertising inventory for customers who want to run advertising across the Amazon advertising network.
- Implemented forecasting for private marketplaces in real-time bidding advertising scenarios.
- Worked with the data science team to operationalize forecasting models.
Founding Principal Engineer
Pani Networks
- Joined the company as the founding engineer, producing initial prototypes for using Romana as a network layer for OpenStack.
- Developed topology-aware IP address management (IPAM) module.
- Produced and shipped editions from PoC/demos to 2.0, up to date.
Founder | CTO
OpenDSP
- Architected, developed, and deployed the initial version of the real-time bidding (RTB) system for bidding on advertising.
- Scaled the system, through architecture and deployment (AWS) to handle hundreds of thousands of qps.
- Led further feature development and operation of the system by managing several teams, including server, UI application, technical ad operations, ops, and DevOps.
- Worked on the company's acquisition by Social Reality.
CTO
Adotube
- Developed initial versions of the company's innovative video ad serving platform.
- Grew the technical team to 30+ people, assuring running ad serving and technical support to customers on a global scale.
- Led the technology aspect of a successful acquisition by Exponential Interactive (now VDX.tv).
Founding Engineer | Team Lead
SnapLogic
- Developed initial version of the GUI, enabling users to create data pipelines.
- Introduced and maintained best engineering practices, such as CI/CD, code coverage, and more.
- Played key roles in shipping releases 0.1 to 3.0 as well as pre-and post-sales support in the company's initial stages.
- Took initiative in prototyping and demoing features I thought would benefit the product, including data pipeline debugger with the ability to step through data flow, and had them included in subsequent versions.
Senior Software Engineer
P&H Solutions
- Worked on porting the Windows-based ASP and C++ web cash management system to the Jakarta EE platform.
- Developed and open-sourced tools for automating similar projects, for ASP and JSP and VBScript and Java conversions.
- Created, working closely with the chief architect, release engineering processes for these new products.
- Developed a reproducible development environment to allow convenient IDE features to work with production-like deployment, thus improving engineers' experience.
Project History
Voitta AI
https://github.com/voitta-ai/voitta-rag is a retrieval service exposed to agents over the Model Context Protocol. It indexes code and document folders, serves semantic and structural search, and keeps a per-agent memory store.
skillz is a shared catalog of agent skills and plugins for Claude Code and Codex, with a schema validator, per-host packaging, and a version-bump gate in CI, so a technique learned once installs everywhere.
voitta-yolt wires LLMs to existing APIs through OpenAPI and MCP, so a tool surface that already exists does not have to be rebuilt for agents.
shmobster is a multi-agent orchestration experiment, written up at https://blog.debedb.com/tag/shmobster/.
The stack is Python, FastAPI, TypeScript, and AWS. Everything is at https://github.com/voitta-ai/.
OpenDSP—Real-time Bidding and Data Management as PaaS
https://blog.debedb.com/tag/opendsp/Without the need to maintain the infrastructure to handle the traffic, this architecture allows users to load the user models into DMP and plugin pricing formulas, as Groovy scripts, and we take care of the rest.
Some details about it are available at the following link: https://blog.debedb.com/tag/opendsp/
Acquired by Social Reality: https://www.prnewswire.com/news-releases/srax-acquires-opendsps-demand-side-platform-300528934.html
Romana Project
https://github.com/romanaThrough Romana's topology-aware IPAM, endpoints receive natively routable addresses: no overlays or tunnels are required, increasing performance and providing operational simplicity. Because IP addresses are assigned with network topology in mind, routes within the network are highly aggregated, reducing the impact on networking hardware and allowing more secure configurations. It supports Kubernetes and OpenStack clusters, on-premise or on AWS.
Wildboard
Dbdb | Cross-Language Debugging Framework
Education
Bachelor's Degree in Computer Science
Massachusetts Institute of Technology - Cambridge, MA, USA
Certifications
Certified Scrum Master
Scrum Alliance
AWS Certified Solutions Architect - Associate
Amazon Web Services (AWS)
Sun Certified Java Developer
Sun Microsystems
Skills
Libraries/APIs
REST APIs, Python API, Asyncio, Apache Lucene, Node.js, Vue 2, Vue, Google Ads API, Facebook SDK, Stripe API, Stripe, Salesforce API, OpenAPI, React
Tools
GitHub, Cron, MySQL Performance Tuning, Eclipse RCP, Git, Jira, LaunchDarkly, Apache Airflow, NGINX, Amazon Athena, AWS Glue, Amazon Elastic MapReduce (EMR), Postman, Amazon Elastic Container Service (ECS), Amazon EKS, AWS Fargate, NPM, Facebook Ads SDK, Amazon Cognito, RabbitMQ, Amazon Elastic Container Registry (ECR), Docker Compose, AWS IAM, Terraform, Claude, Istio, Grafana, Claude Code
Languages
Java, Python, SQL, Go, PHP, Scala, Lua, Groovy, C, Active Server Pages (ASP), C++, Visual Basic, Python 3, JavaScript, TypeScript
Paradigms
ETL, Back-end Architecture, Test-driven Development (TDD), Lambda Architecture, DevOps, Microservices, Scrum, Agile, Business Intelligence (BI), Management, REST, Dependency Injection, Model Context Protocol (MCP)
Platforms
Amazon EC2, Amazon Web Services (AWS), Java EE (Jakarta EE), Linux, Algolia, Kubernetes, Docker, Apache Pig, AWS Lambda, OpenStack, Oracle, Eclipse, Android, Heroku, Google Ads, Google Cloud Platform (GCP), Firebase, AWS ALB
Storage
MySQL, Databases, JSON, NoSQL, Relational Databases, Cloud Deployment, PostgreSQL, MySQL Server, Vertica, Redis, Amazon DynamoDB, CouchDB, Redshift, Amazon S3 (AWS S3), MongoDB, Google Cloud Datastore, Google Cloud, Cloud Firestore, CockroachDB, IBM Db2
Frameworks
Spring, Swagger, Spark, Hadoop, Adobe Flex, Jakarta Server Pages (JSP), Spring Boot, PhoneGap, Next.js, OAuth 2, Flask, NestJS, Chainlit, DSPy
Other
Software Engineering, HTTP REST, APIs, AWS Cloud Architecture, Architecture, Cloud, CSV, Scalable Architecture, Software Design, Cloud Architecture, Software Architecture, Distributed Systems, Back-end, Orchestration, Data Feeds, Technical Architecture, API Integration, Web Servers, System Architecture, CTO, AWS Certified Solution Architect, Team Leadership, Technical Leadership, CI/CD Pipelines, Amazon RDS, Algorithms, Fractional CTO, SaaS, AI Integration, Computer Science, Scrum Master, Site Reliability Engineering (SRE), Streaming Data, Data Engineering, Windows 10, Advertising, Windows 10 Touchscreen, Touchscreens, Digital Kiosks, Cordova, Facebook Ads, Advertising Technology (Adtech), FastAPI, IT, IT Project Management, Lambda Functions, Solution Architecture, OAuth, Product Management, Payment APIs, OpenAI, LangChain, Anthropic, Agentic AI, AI Agents, Large Language Models (LLMs), Prometheus, Retrieval-augmented Generation (RAG), Artificial Intelligence (AI), Cloud Migration
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring