Hernan is currently unavailable
Hernan Marano
Verified Expert in Engineering
Senior Full-stack Developer
Buenos Aires, Argentina
Toptal member since January 8, 2016
Bio
Hernán is an AI engineer and software architect with 23 years of experience who turns unclear problems into AI systems that thrive in production. He builds LLM applications and the evaluation and observability layer around them, sets technical direction, and stays in the code. Most of his career has been on multi-tenant platforms in healthcare, insurance, and legal, where failure is expensive. He is completing an M.Sc. in artificial intelligence.
Portfolio
DealDecision AI
AI Agents, Large Language Models (LLMs), Next.js, Supabase...
Craig.AI
C#, Entity Framework, SQL, Large Language Models (LLMs)...
Dymeng Technology Solutions (acquired by ATrack Solutions)
Architecture, .NET Core, .NET Framework, SQL, CI/CD Pipelines...
Experience
- C# - 18 years
- .NET - 18 years
- Software Architecture - 10 years
- TypeScript - 5 years
- Azure - 4 years
- Python - 2 years
- Large Language Model Operations (LLMOps) - 1 year
- Large Language Models (LLMs) - 1 year
Work Experience
Technical Advisor
2026 - PRESENT
DealDecision AI
- Advised the CTO of an LLM-powered deal-analysis platform for VC and PE investors on what to build and in what order, preparing it for launch.
- Turned an open-ended roadmap document into a phased functional backlog with a dependency map and an open-decisions log.
- Refactored the platform to close the security and logging gaps that blocked the production launch.
- Reviewed another developer's pull requests to keep the codebase release-ready.
Technologies: AI Agents, Large Language Models (LLMs), Next.js, Supabase, Technical Leadership, Product Strategy, TypeScript
Lead Engineer
2025 - PRESENT
Craig.AI
- Built the LLM pipeline that generates clinical reports from clinical data, and the evaluation and observability layer around it.
- Traced every model call with OpenTelemetry into Langfuse, with cost, latency, and quality visible per prompt version.
- Gated every change against a golden set of synthetic, real, and adversarial cases and an LLM-as-a-judge being calibrated with 500+ structured clinician annotations.
- Worked with the clinical and product side to define what a correct report looks like and turned that definition into evaluation criteria.
- Owned the platform end to end: a .NET and React application on Azure, infrastructure as code with isolated environments, and CI/CD.
- Designed the multi-tenant administration service that handles tenant configuration, identity providers, feature flags, and the catalog of AI vendors.
Technologies: C#, Entity Framework, SQL, Large Language Models (LLMs), Large Language Model Operations (LLMOps), Prompt Engineering, .NET, React, Terraform, GraphQL, Software Architecture, Azure, TypeScript, OpenTelemetry
Principal Solutions Engineer
2019 - 2025
Dymeng Technology Solutions (acquired by ATrack Solutions)
- Joined a three-person company as its full-time engineer and stayed as it grew into a delivery team of four to five developers, a tester, and a scrum master.
- Set the architecture and technical direction for a healthcare platform serving ABA clinics, used by thousands of patients and hundreds of providers.
- Led the transition from a monolith to message-bus-connected services and designed the JWT-based identity and authorization model.
- Built the CI/CD pipeline covering web deployments to AWS, app-store releases, and database migrations, and supported the migration from .NET Framework to .NET Core.
- Hired, onboarded, and evaluated developers, and took part in client meetings where product decisions were made.
Technologies: Architecture, .NET Core, .NET Framework, SQL, CI/CD Pipelines, Amazon Web Services (AWS), Microsoft SQL Server, Elasticsearch, Vue, JavaScript, SQL Server Integration Services (SSIS), Jenkins, Domain-driven Design (DDD), .NET, Azure SQL Databases, ASP.NET Core, Identity, ASP.NET Identity, JSON Web Tokens (JWT), OpenID, OAuth 2, Authentication, Authorization, Single Sign-on (SSO), IIS, API Integration, Technical Leadership, Entity Framework, Cloud, Code Review, AI Tools, User Experience (UX), User Interface (UI), API Design, System Integration, HTML, Web Development, Algorithms, Software Architecture, Automated Testing, Object-oriented Programming (OOP), Entity Framework Core, SOLID Principles, Event-driven Design (EDD), Swagger, Data Modeling, Cloud Deployment, ESLint, Redis, Amazon S3 (AWS S3), Microservices Architecture, SSL, HTTPS, REST, Hangfire, JSON, ASP.NET MVC 5, Microsoft Visual Studio, C#, ASP.NET MVC, Windows, Git, ASP.NET Web API, Dependency Injection, Full-stack, Front-end, Software Engineering, Databases, Back-end, Database Administration (DBA), CSS3, Razor Pages, RabbitMQ, Visual Studio Code (VS Code), Multi-tenant SaaS, Distributed Systems, TypeScript
Full-stack Engineer
2017 - 2019
Dymeng Services, Inc.
- Contributed as a full-stack developer across multiple web applications, focusing on both feature development and long-term architectural improvements.
- Charged with maintaining and expanding functionality, integrating with third-party APIs, and improving overall system reliability.
- Drafted the plan to move off the legacy monolith toward services, aiming at scalability, maintainability, and team velocity — the work he then led from 2019.
Technologies: C#, ASP.NET MVC 5, JavaScript, HTML5, Microsoft SQL Server, CSS3, Razor Pages, Event-driven Architecture, Amazon Web Services (AWS), AWS Lambda, GraphQL, RabbitMQ, Vue, Domain-driven Design (DDD), .NET, Azure SQL Databases, Docker, ASP.NET Core, Identity, ASP.NET Identity, JSON Web Tokens (JWT), OpenID, OAuth 2, Authentication, Authorization, Single Sign-on (SSO), IIS, Angular, Microsoft Entra ID, Architecture, API Integration, Technical Leadership, Entity Framework, Cloud, CI/CD Pipelines, Code Review, User Experience (UX), User Interface (UI), API Design, System Integration, CSS, HTML, Web Development, Algorithms, Software Architecture, Automated Testing, NUnit, Object-oriented Programming (OOP), Entity Framework Core, SOLID Principles, Event-driven Design (EDD), Swagger, Data Modeling, Cloud Deployment, ESLint, NPM, Redis, Amazon S3 (AWS S3), Microservices Architecture, SSL, HTTPS, REST, Hangfire, Model View ViewModel (MVVM), Groovy, JSON, Microsoft Visual Studio, SQL, ASP.NET MVC, Windows, jQuery, Git, ASP.NET Web API, Dependency Injection, Full-stack, Front-end, Software Engineering, Databases, Back-end, Database Administration (DBA), .NET Framework, SQL Server Integration Services (SSIS), Visual Studio Code (VS Code), Multi-tenant SaaS
Full-stack Engineer
2017 - 2018
Fieldcraft, Inc.
- Refactored the legacy application to isolate and abstract components, reorganized module communication, and improved general architecture.
- Combed the applications to remove bugs, improve performance, and polish UI components.
- Improved asset management by building an uploader UI component and back components for storing and indexing.
Technologies: Knockout (Knockout.js), ASP.NET MVC, .NET, Bootstrap, jQuery, Microsoft Visual Studio, Microsoft SQL Server, C#, JavaScript, Git, GitHub, Dependency Injection, Full-stack, ASP.NET, Event-driven Architecture, Google Analytics, IIS, API Integration, Cloud, User Experience (UX), User Interface (UI), CSS, HTML, Web Development, Algorithms, Object-oriented Programming (OOP), SOLID Principles, Data Modeling, SSL, HTTPS, REST, JSON, SQL, Windows, Front-end, Software Engineering, Databases, Back-end, Database Administration (DBA), CSS3, REST APIs
Full-stack Engineer
2016 - 2017
ICVM Group
- Maintained and extended a multi-tenant CMS serving around a hundred law firms: new features, performance work, and stability across client sites.
- Extended the architecture so white-label sites could be customized through configuration alone, balancing per-client customization against a single maintainable core.
- Extended the data model to support new features, dynamic components, and new content types.
Technologies: SQL, JavaScript, jQuery, Windows Server, Adobe ColdFusion, Regex, TortoiseSVN, Database Administration (DBA), IIS, Architecture, Windows, C#, Git, Dependency Injection, Full-stack, Google Analytics, Node.js, Cloud, User Experience (UX), User Interface (UI), CSS, HTML, Web Development, Algorithms, Object-oriented Programming (OOP), SOLID Principles, Data Modeling, Bootstrap, SSL, HTTPS, Model View ViewModel (MVVM), JSON, Microsoft SQL Server, Front-end, Software Engineering, Databases, Back-end, CSS3
Founder & Lead Developer
2003 - 2015
EQIP
- Founded and ran a software company serving the event-management industry for 12 years, owning client conversations, requirements, product design, and architecture with a small team.
- Developed a PDF manipulation module that converted PDF files into an array of images, extracted text, and indexed files and their contents using Elasticsearch. Files could be browsed and displayed with a web-based interactive viewer.
- Designed and implemented a complex attendee registration flow that involved the integration of many online payment gateway APIs, like POST-based, REST, WCF, and more.
- Wrote a script that migrated information from a complex multi-table SQL database into a document-based NoSQL database. The task involved maintaining reference constraints, reorganizing data, and calculating the aggregates.
- Created a real-time social media aggregator with an integrated custom moderation engine and a context-sensitive sorting algorithm to identify the most relevant social media posts.
- Supported a task scheduler to initialize, schedule, run, and cancel jobs. Provided these services to other applications and users alike.
Technologies: SQL, SignalR, Knockout (Knockout.js), JavaScript, ASP.NET Web API, ASP.NET MVC, C#, .NET, Windows, Microsoft Visual Studio, jQuery, XML, Git, Sass, Elasticsearch, MongoDB, XSLT, Unit Testing, Dependency Injection, Parallel Computing, Hangfire, REST, GhostScript, Apache Solr, Full-stack, ASP.NET Web Forms, Azure, Python, Document Processing, IIS, Entity Framework, Cloud, User Interface (UI), CSS, HTML, Web Development, Algorithms, NUnit, Object-oriented Programming (OOP), SOLID Principles, Data Modeling, Bootstrap, SSL, HTTPS, X (formerly Twitter) API, Model View ViewModel (MVVM), JSON, ASP.NET MVC 5, Adobe ColdFusion, Microsoft SQL Server, Less, GitHub, Front-end, Software Engineering, Databases, Back-end, Database Administration (DBA), CSS3, Razor Pages, .NET Framework, REST APIs
Full-stack Web Developer
2012 - 2012
Inworx/Smartix
- Designed the architecture and coded a CMS for a multi-tenant white-label POS. It involved creating a web app that displayed the stored contents and the content administrator and editor, featuring a wide array of custom controls.
- Implemented a mechanism for a white-label website by parsing and caching LESS files that had been previously filled in with customer-defined variables.
- Implemented a common asset storage and retrieval mechanism. These assets consisted of images, documents, and plain or HTML-formatted text.
- Improved the user experience by keeping the user aware of their progress. The polling mechanism was replaced with a system in which the server pushed notifications to the client.
Technologies: Knockout (Knockout.js), SignalR, Less, ASP.NET MVC, Full-stack, Front-end, Microsoft SQL Server, XML, Model View Controller (MVC), JSON, Windows, C#, SQL, Microsoft Visual Studio, JavaScript, jQuery, Git, Sass, ASP.NET Web API, Unit Testing, Dependency Injection, Parallel Computing, REST, ASP.NET, Google Analytics, .NET, IIS, API Integration, Entity Framework, Cloud, User Interface (UI), System Integration, CSS, HTML, Web Development, Algorithms, NUnit, Object-oriented Programming (OOP), SOLID Principles, Data Modeling, Bootstrap, SSL, HTTPS, Model View ViewModel (MVVM), ASP.NET MVC 5, Software Engineering, Databases, Back-end, Database Administration (DBA), CSS3, Razor Pages, .NET Framework, REST APIs
Project History
M.Sc. Thesis: Evidence Beats Anecdote — Evaluation and Monitoring of an LLM Agent
Master's thesis at Universidad de San Andrés, in progress and expected in December 2026. Full title: "Evidence Beats Anecdote: Incremental Evaluation and Monitoring of an LLM Agent Under Resource Constraints (An Instrumented Case Study)".
An instrumented case study on keeping an LLM agent trustworthy in production when the team has no dedicated evaluation budget: tracing every model call with OpenTelemetry, building a golden set from synthetic, real, and adversarial cases, and calibrating an LLM-as-a-judge against human annotation so that quality, cost, and latency are measured per prompt version, and regressions are caught before a user sees them. It also compares monitoring platforms across scenarios, prompts, and rubrics, and sets out an incremental adoption path for small teams.
The question behind it is the one I ask of every system: how do you know it's working?
An instrumented case study on keeping an LLM agent trustworthy in production when the team has no dedicated evaluation budget: tracing every model call with OpenTelemetry, building a golden set from synthetic, real, and adversarial cases, and calibrating an LLM-as-a-judge against human annotation so that quality, cost, and latency are measured per prompt version, and regressions are caught before a user sees them. It also compares monitoring platforms across scenarios, prompts, and rubrics, and sets out an incremental adoption path for small teams.
The question behind it is the one I ask of every system: how do you know it's working?
Hydrocarbon Production Forecasting — MLOps Platform
https://github.com/hernangm/mia204_01An end-to-end ML engineering platform that forecasts monthly gas production for unconventional wells from Argentina's public energy data, built for a graduate course in AI in Production. Apache Airflow orchestrates a monthly pipeline: downloading, preprocessing, building a PostgreSQL feature store with point-in-time correctness, training four model variants, and promoting the lowest-RMSE run to the production alias in MLflow. A separate FastAPI service serves forecasts, Ray Serve handles inference, and a weekly drift DAG runs PSI and KS tests per feature, plus RMSE-decay checks and triggers retraining. I designed the architecture and built the pipeline, feature store, and API with a classmate; the emphasis was reproducibility, traceability, and automation rather than model accuracy.
API (REST and GraphQL)
https://github.com/hernangm/portfolio-dotnet-apiI developed a comprehensive .NET 9 API to demonstrate modern software engineering practices and architectural patterns. The project, structured as a modular monolith, provides a robust backend for managing academic abstracts and features secure, JWT-based user authentication. It uniquely offers dual API access through both traditional RESTful endpoints and a flexible GraphQL API, which leverages CQRS principles to separate read and write operations. The tech stack includes ASP.NET Core, Entity Framework Core with a custom interceptor for automated entity auditing, and MSTest for unit and integration testing. The entire development lifecycle is automated with a full CI/CD pipeline using GitHub Actions, which handles everything from building and testing to deploying the application to Azure and running smoke tests against the live environment to verify a successful rollout.
Identity Server
https://github.com/hernangm/portfolio-dotnet-identityPortfolio.Dotnet.Identity is a .NET 9 Identity Server for auth, supporting user management, local & external providers (Azure AD, Google), & role-based access. It uses IdentityServer4, EF Core, & ASP.NET Core Razor Pages, with SQL Server for storage, email, & Swagger API docs.
Education
2025 - 2026
Master of Artificial Intelligence Degree in Artificial Intelligence
Universidad de San Andres - Buenos Aires, Argentina
2012 - 2012
Master in User Experience, Usability and Accessibility in Usability, User Experience, Accessibility
Universidad Tecnologica Nacional - Buenos Aires, Argentina
2003 - 2010
Bachelor's Degree in Computer Science
Universidad Centro de Altos Estudios en Ciencias Exactas - Buenos Aires, Argentina
Certifications
SEPTEMBER 2025 - PRESENT
Unsupervised Learning, Recommenders, Reinforcement Learning
Stanford University | via Coursera
SEPTEMBER 2025 - PRESENT
Machine Learning
Stanford University | via Coursera
AUGUST 2025 - PRESENT
Advanced Learning Algorithms
Stanford University | via Coursera
JULY 2025 - PRESENT
Supervised Machine Learning: Regression and Classification
Stanford University | via Coursera
JANUARY 2023 - PRESENT
Certificate of Specialization in Web Usability and Accessibility
Universidad Tecnológica Nacional (U.T.N.)
Skills
Libraries/APIs
jQuery, Entity Framework, REST APIs, SignalR, X (formerly Twitter) API, Vue, OpenID, Node.js, NumPy, Scikit-learn, React, PyTorch, TensorFlow
Tools
Microsoft Visual Studio, Git, GitHub, Hangfire, RabbitMQ, Jenkins, ChatGPT, Terraform, Apache Solr, TortoiseSVN, Google Analytics, NPM, Apache Airflow
Languages
SQL, XML, JavaScript, C#, HTML5, GraphQL, HTML, Sass, Less, XSLT, Groovy, Regex, CSS3, Python, CSS, TypeScript, GhostScript
Frameworks
Knockout (Knockout.js), ASP.NET Web API, ASP.NET MVC, .NET, ASP.NET, .NET Core, ASP.NET MVC 5, Razor Pages, ASP.NET Web Forms, ASP.NET Core, ASP.NET Identity, JSON Web Tokens (JWT), Entity Framework Core, Swagger, NUnit, Bootstrap, OAuth 2, Next.js, Angular, Tailwind CSS
Paradigms
Model View Controller (MVC), REST, Event-driven Architecture, Object-oriented Programming (OOP), Event-driven Design (EDD), Microservices Architecture, Model View ViewModel (MVVM), Unit Testing, Dependency Injection, Parallel Computing, Automated Testing
Platforms
Adobe ColdFusion, Amazon Web Services (AWS), Visual Studio Code (VS Code), Windows, Windows Server, Azure, AWS Lambda, Docker
Storage
Microsoft SQL Server, JSON, Databases, Elasticsearch, Database Administration (DBA), SQL Server Integration Services (SSIS), Azure SQL Databases, Cloud Deployment, Redis, Amazon S3 (AWS S3), MongoDB, Microsoft Entra ID, PostgreSQL
Other
Full-stack, Software Engineering, Back-end, Architecture, .NET Framework, Domain-driven Design (DDD), Identity, Authentication, Authorization, Single Sign-on (SSO), API Integration, Technical Leadership, Cloud, User Interface (UI), API Design, System Integration, Web Development, Algorithms, Software Architecture, SOLID Principles, Data Modeling, Multi-tenant SaaS, IIS, Front-end, Usability, CI/CD Pipelines, OpenAI, Document Processing, Artificial Intelligence (AI), Large Language Models (LLMs), Code Review, AI Tools, User Experience (UX), Identity Server, ESLint, SSL, HTTPS, Large Language Model Operations (LLMOps), Prompt Engineering, AI Agents, Supabase, Product Strategy, Machine Learning Operations (MLOps), FastAPI, Distributed Systems, OpenTelemetry, Accessibility, Machine Learning, Neural Networks, Supervised Learning, Unsupervised Learning, Reinforcement Learning, Recommendation Systems, Regression, AI Integration, MLflow
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring