Mircea is available for hire
Hire MirceaMircea Ene
Verified Expert in Engineering
Software Developer
Bucharest, Romania
Toptal member since October 4, 2021
Bio
Mircea is an AI Engineer building voice-first, LLM-driven products end-to-end. He architects multi-agent systems with Pydantic AI, AutoGen, LangGraph, and LangChain, fine-tunes Mistral models, and builds LLM-as-a-judge evaluation pipelines across GPT, Claude, Mistral, and Grok. He's shipped real-time voice pipelines with Deepgram and Whisper, and cut server load by 75% through smart caching. Mircea has four years of C++ systems development work at Electronic Arts and a $10,000 ElevenLabs grant.
Portfolio
Healthcare Payments Partners, Inc.
AI Agents, AI Agent Orchestration, Amazon Web Services (AWS), Python, Node.js...
Augmented Data Management Solutions Inc.
Twilio, Architecture, n8n, Agentic AI, ElevenLabs Solutions, ChatGPT, Figma...
Considr Inc.
Node.js, TypeScript, Architecture, Google Cloud Platform (GCP), Socket.IO...
Experience
- Node.js - 5 years
- Debugging - 5 years
- C++ - 5 years
- Python - 4 years
- Windows - 4 years
- C# - 3 years
- JavaScript - 3 years
- MongoDB - 2 years
Work Experience
[FT/PT] [AI Engineer] for a [Consumer Healthcare AI Platform]
2026 - PRESENT
Healthcare Payments Partners, Inc.
- Designed the outbound AI voice agent (Retell) and its prompts: calls patients at booked times, fills enrollment gaps, quotes M3P costs and texts e-sign links mid-call.
- Built the Telnyx voice stack: inbound AI intake agent with tool calls and warm human handoff, browser conference calls, IVR with voicemail, real-time Deepgram STT.
- Architected the AWS platform in CDK (7-account org): FastAPI + 5 React apps on ECS Fargate, DynamoDB, S3, KMS, SES, EventBridge, GitHub OIDC CI/CD.
- Engineered HIPAA controls: transactional PHI-access audit trail on DynamoDB, KMS encryption, OTP-gated patient sessions, malware-scanned uploads.
- Integrated Medicare data: CMS Blue Button claims (OAuth/PKCE), monthly ingest of ~7,600 plans and Part D formularies, X12 270/271 eligibility, M3P cost projections.
Technologies: AI Agents, AI Agent Orchestration, Amazon Web Services (AWS), Python, Node.js, React, Text-to-Speech (TTS), Speech-to-Text (STT), LLM Integration, HIPAA, AI Voice Agents, Amazon Bedrock, Amazon Cognito, Testing, Software Development, Amazon Elastic Container Service (ECS), AI-assisted Development, Codex, Monorepos, Bug Fixes, Agentic Frameworks, Retrieval-augmented Generation (RAG), Agentic AI Systems, RAG Architecture, OAuth, Identity & Access Management (IAM), OpenAI SDK, API Development, Agentic Coding, Multi-tenant SaaS, Cloud, Application Security, AI Engineering, Data Science, Gemini, Docker, Automated Testing, AI Inference, Front-end, Dashboards, Security, Plaid, AI Development, AI Evaluation, Solution Architecture, Data Engineering, Scaling, HIPAA Compliance, CTO, Cursor AI, Firebase, Technical Leadership, Linear, Project Management, Communication, Go
AI Consultant
2026 - PRESENT
Augmented Data Management Solutions Inc.
- Migrated the voice pipeline from ElevenLabs to Amazon Connect, consolidating Twilio + ElevenLabs into one AWS-native stack, unlocking HIPAA infrastructure, human-agent escalation, and Contact Lens analytics — at the cost of slightly less natural TTS.
- Shipped a modular React and Vite platform serving four role-scoped portals plus a marketing site from one app, using lazy-loaded route sections for independent bundles and isolated auth contexts to enforce role-based access against a shared library.
- Built a real-time tool-call pipeline for an LLM voice agent with a pre-result hook that rebuilds and pushes the system prompt to the model between a tool's execution and its result delivery, keeping context coherent across mid-call state mutations.
- Engineered a distributed booking-concurrency model using per-time-bucket write locks inside MongoDB transactions. Overlapping bookings serialize and retry on fresh snapshots while non-overlapping ones commit in parallel, preventing double-booking.
- Architected a multi-tenant telephony SaaS that routes inbound calls by dialed number through a cached tenant resolver, enforces subscription state at the TwiML layer, and isolates per-tenant phone numbers, voice config, and SMTP under shared infra.
- Designed a pluggable voice-agent abstraction unifying ElevenLabs and Deepgram behind one interface, normalizing wire formats, tool protocols, and transcript events so the active provider hot-swaps per tenant or per call from config alone.
Technologies: Twilio, Architecture, n8n, Agentic AI, ElevenLabs Solutions, ChatGPT, Figma, Figma Make, Retrieval-augmented Generation (RAG), Agentic RAG Systems, Claude Code, Full-stack, AI Chatbots, AI Prompts, Anthropic, Chatbots, Large Language Models (LLMs), Artificial Intelligence (AI), Claude API, Healthcare, Payments, Supabase, Vercel, Software Architecture, Software Design, NestJS, Amazon Connect, Scalability, RAG Systems, Push Notifications, Generative Artificial Intelligence (GenAI), Leadership, API Backwards Compatibility, AI Agent Orchestration, LLM Integration, Text-to-Speech (TTS), Speech-to-Text (STT), UI Design, AI Voice Agents, Testing, Software Development, Amazon Elastic Container Service (ECS), AI-assisted Development, Codex, Monorepos, Bug Fixes, Agentic Frameworks, Agentic AI Systems, RAG Architecture, OAuth, Identity & Access Management (IAM), OpenAI SDK, API Development, Agentic Coding, Multi-tenant SaaS, Cloud, Application Security, AI Engineering, Data Science, Gemini, Docker, Automated Testing, AI Inference, Front-end, Dashboards, Security, AI Development, AI Evaluation, Solution Architecture, Data Engineering, Concurrency, HIPAA Compliance, CTO, Cursor AI, Firebase, Technical Leadership, Linear, Project Management, Communication, SQL
Founding Engineer
2022 - PRESENT
Considr Inc.
- Architected a Node.js server that handles thousands of concurrent web sockets transferring real-time audio data from browser clients.
- Architected an agentic framework using Pydantic AI and a voice agent using Deepgram.
- Structured and pruned big data that was used to fine-tune a custom LLM model based on Mistral-7B using data gathered from users.
- Configured a MERN stack application on GCP using features like Cloud Run, VPN, buckets (storage), and authentication.
- Took features from brainstorming ideas and made them fully functional.
- Integrated several APIs from providers like GPT, Stability.AI, ElevenLabs, Stripe, and Deepgram.
Technologies: Node.js, TypeScript, Architecture, Google Cloud Platform (GCP), Socket.IO, MongoDB, React, React Native, Performance, Full-stack, Azure, Next.js, OpenAI API, Prompt Engineering, API Architecture, API Design, API Testing, Amazon S3 (AWS S3), Cloud Architecture, Cloud Storage, FastAPI, Web Applications, Databases, SQL, Claude, OpenAI GPT-3 API, OpenAI GPT-4 API, Audio Recording, Expo, Web Development, Large Language Models (LLMs), Artificial Intelligence (AI), AI Agents, Startups, Asynchronous Programming, MongoDB Atlas, ChatGPT, OpenAI, JSON Web Tokens (JWT), Webhooks, SendGrid, Google Analytics, Back-end Architecture, Cloud Infrastructure, Serverless Architecture, Amazon Web Services (AWS), Full-stack Development, AI Integration, Agentic AI, Agentic Frameworks, LangChain, LangGraph, n8n, Scripting, Python Script, Pydantic, Model Context Protocol (MCP), Coding, Android, App Store, Minimum Viable Product (MVP), SaaS, Stripe Terminal, User Interface (UI), iOS, Stripe Connect, System Architecture, Figma, Figma Make, Retrieval-augmented Generation (RAG), Agentic RAG Systems, Amazon Connect, AI Chatbots, AI Prompts, Anthropic, Chatbots, Claude API, Payments, Vercel, Software Architecture, Software Design, NestJS, Scalability, RAG Systems, Push Notifications, Generative Artificial Intelligence (GenAI), Leadership, API Backwards Compatibility, AI Agent Orchestration, LLM Integration, Text-to-Speech (TTS), Systems Engineering, UI Design, Software Development, Monorepos, Agentic AI Systems, RAG Architecture, OAuth, Identity & Access Management (IAM), API Development, Cloud, AI Engineering, Gemini, Docker, Automated Testing, AI Inference, Front-end, Dashboards, SOC 2, Security, AI Development, AI Evaluation, Solution Architecture, Data Engineering, Scaling, CTO, Cursor AI, Firebase, Technical Leadership, Project Management, Communication
Senior Software Engineer
2021 - 2022
Latch
- Developed back-end features using Java, Spring Boot, and Play Framework to communicate with embedded devices via FlatBuffers.
- Tracked and fixed bugs using Jira as a reporting tool.
- Developed scripts to automate certain time-consuming activities and generate code for the FlatBuffers integration with Java.
Technologies: Amazon Web Services (AWS), Java 8, Spring Boot, FlatBuffers, Protobuf, gRPC, REST, Play Framework, API Integration, REST APIs, JSON, XML, APIs, Multithreading, Spring, Stripe, Back-end, Back-end Development, AWS Lambda, GitHub, GraphQL, PostgreSQL, Amazon RDS, Microservices, Azure, API Architecture, API Design, API Testing, Databases, Web Development, Asynchronous Programming, MongoDB Atlas, Webhooks, Back-end Architecture, Cloud Infrastructure, Kubernetes, Serverless Architecture, System Architecture, Software Architecture, Software Design, Push Notifications, API Backwards Compatibility, SSH, Testing, Software Development, Monorepos, OAuth, Identity & Access Management (IAM), API Development, Multi-tenant SaaS, Cloud, Application Security, Data Science, Docker, Automated Testing, Front-end, Dashboards, Security, Solution Architecture, Concurrency, Scaling, Communication, SQL, Go
Senior Software Engineer
2021 - 2022
Aptiv
- Implemented a CI/CD pipeline for a couple of projects that were missing it entirely and added VectorCAST as a unit test framework to the projects.
- Investigated and identified problems in various project development processes using different build tools like CMake, Make, Bazel, and MSBuild. Used various compilers like TASKING, IAR, Microsoft Visual Studio, GCC, MinGW, and GNU.
- Developed a framework using Docker, JFrog Artifactory, and Jenkins to allow teams to export their environment as a Docker image later deployed on different AWS and GCP machines.
Technologies: C++, Jenkins, Docker, Amazon Web Services (AWS), JFrog, CloudBees, CMake, Make (formerly Integromat), Ant Design, MinGW, GNU, Microsoft Visual Studio, JSON, XML, APIs, C++17, Multithreading, Linux, Architecture, Performance, Microservices, CI/CD Pipelines, DevOps, API Architecture, Databases, Web Development, Webhooks, Back-end Architecture, Cloud Infrastructure, Serverless Architecture, System Architecture, Software Architecture, API Backwards Compatibility, SSH, Systems Engineering, Testing, Software Development, Monorepos, OAuth, API Development, Real Estate, Technical Leadership, Project Management, Communication
Software Engineer
2017 - 2021
Electronic Arts
- Developed a variation of the Hungarian algorithm to replace a greedy algorithm.
- Created a C# tool for merging glyphs from different fonts into a new font. This was useful in scenarios where various languages are used, and complications might appear when two glyphs from two different languages share the identical Unicode.
- Developed a C# tool using the Selenium framework that would complete the end-to-end process of downloading, updating, and generating C++ code and uploading configuration files to web apps that wouldn't support REST. This also included authentication.
- Developed a C# tool that would replace manually updating XML files for configuring Flash and ActionScript menu UIs with a visual UI that would later convert to XML.
- Delivered a context-switch-prone environment helping other teams reach their deadlines.
Technologies: C++, C#, Python, Java, JSON, XML, APIs, C++17, Multithreading, Architecture, Performance, .NET, Back-end, Back-end Development, GitHub, Data Scraping, Microservices, Microsoft SQL Server, .NET Core, CI/CD Pipelines, API Design, API Testing, Databases, Web Development, Back-end Architecture, Cloud Infrastructure, User Interface (UI), System Architecture, Software Architecture, Software Design, Scalability, API Backwards Compatibility, Systems Engineering, Testing, Software Development, Monorepos, Bug Fixes, OAuth, API Development, Front-end, Dashboards, Concurrency, Project Management, Communication
Software Engineer
2018 - 2018
Freelance
- Developed a scalable C# application based on the Selenium framework that would automatically complete the end-to-end process of purchasing tickets from a website using proxies and captchas, as well as solving features and API rate limit avoidance.
- Implemented a socket communication module between C# on the front end and Java on the back end.
- Created and trained a machine learning model in Python, using mainly NumPy and Pandas modules for solving image matching captchas.
- Developed a web-based dashboard for controlling hundreds of instances running on Windows Server 2012 machines.
- Implemented notifications for events like "ticket found" and "ticket purchased" and errors, using the Twilio SMS framework.
Technologies: Java, C#, Selenium, Python, Machine Learning, API Integration, REST APIs, JSON, XML, APIs, Multithreading, Linux, Architecture, .NET, Back-end, Back-end Development, GraphQL, WebSockets, Data Scraping, React, Express.js, .NET Core, Next.js, API Architecture, API Design, API Testing, Web Applications, Databases, Back-end Architecture, Supabase, Vercel, Software Design, Software Development, Monorepos, Firebase, Communication
Software Engineer
2016 - 2017
Simultec
- Developed an OpenGL application to render a 3D map based on 2D cartographic images and Elevation API from Google.
- Crafted a 2D OpenGL UI that would replicate an airplane's cockpit based on images cropped by me in CorelDRAW of the actual cockpit instruments.
- Developed a C++ cloud-like console application using sockets that would allow clients to download and upload files to the server.
- Created a module for importing Blender models into OpenGL.
Technologies: C++, OpenGL, CorelDRAW, Windows, Socket.IO, Databases, Software Development, Monorepos, Communication
Project History
Hollywood Trivia Game
I created a Node.js back end, and a Unity front-end mobile application. I used Atlas MongoDB for permanent storage and Redis as an in-memory cache. The client communicated via REST and requested with the Node.js Express server. In order to increase throughput, all the puzzles in the game are pre-built and stored in the DB. I also developed a backtracking algorithm combined with dynamic programming and used a scoring mechanism based on multiple attributes of a movie to create the puzzles. The data used to construct the puzzles is scraped daily for updates (new movies and new actors) using GET requests to API endpoints provided by The Movie Database (TMDB) for free.
The difficulty of a puzzle is dynamically calculated based on a machine learning model trained using user feedback and movie preferences. For generating puzzles, I chose to go with C++ worker threads spawned by Node.js V8 engine. This solution was based on having all the data structures already written in JavaScript. For unit, integration, and E2E tests I used Jest, a testing framework available as a Node.js package. The servers are run on Linux machines hosted on GCP.
The difficulty of a puzzle is dynamically calculated based on a machine learning model trained using user feedback and movie preferences. For generating puzzles, I chose to go with C++ worker threads spawned by Node.js V8 engine. This solution was based on having all the data structures already written in JavaScript. For unit, integration, and E2E tests I used Jest, a testing framework available as a Node.js package. The servers are run on Linux machines hosted on GCP.
Dashboard for Sensor Statistics
I created a raw HTML dashboard backed by a PHP application used for monitoring different parameters of air quality, humidity, and UV levels. The application runs in an Apache environment and uses MySQL as a database.
Using the Stripe payment module, I implemented a payment feature. Also, I implemented an admin role and user management.
Using the Stripe payment module, I implemented a payment feature. Also, I implemented an admin role and user management.
Founding Engineer
https://app.vnote.aiA speech-to-text application that uses a custom-trained LLM to organize ideas into a desired document format. The app further allows you to improve the generated writing with features like versioning/feedback requests from others and text-to-speech.
Footbal World Cup 2026 Predictions
https://mondy.footballA React app where you can create groups and invite friends to compete against with predictions for the FIFA World Cup 2026. I was the sole engineer for this project and handled everything from design to CI/CD pipeline and marketing.
Education
2013 - 2017
Bachelor's Degree in Systems Engineering
Politehnica University of Bucharest - Bucharest, Romania
Skills
Libraries/APIs
Node.js, REST APIs, React, OpenAI API, Claude API, API Development, Stripe, Socket.IO, Auth0 API, Auth0 Management API v2, Pydantic, Stripe Connect, OpenGL, Stripe API, FlatBuffers, Protobuf
Tools
Claude, ChatGPT, MongoDB Atlas, AI Prompts, GitHub, Expo, SendGrid, Google Analytics, n8n, Figma, Claude Code, Amazon Elastic Container Service (ECS), Codex, Plaid, CorelDRAW, Jenkins, CMake, Make (formerly Integromat), Apache, MinGW, Microsoft Visual Studio, Auth0, Amazon Connect, Amazon Cognito
Languages
C++, JavaScript, Python, Java, XML, TypeScript, SQL, Python Script, Go, C#, C++17, GraphQL, HTML5, PHP 7, Java 8, HTML, CSS, TypeScript 5
Frameworks
Express.js, React Native, JSON Web Tokens (JWT), Agentic Frameworks, LangGraph, NestJS, OAuth 2, Spring, .NET, .NET Core, Next.js, Django, Selenium, Ant Design, Spring Boot, gRPC, Play Framework
Paradigms
REST, API Architecture, Asynchronous Programming, Back-end Architecture, Microservices, Serverless Architecture, Model Context Protocol (MCP), UI Design, Testing, Automated Testing, HIPAA Compliance, DevOps
Platforms
Amazon Web Services (AWS), Vercel, Windows, Visual Studio Code (VS Code), Google Cloud Platform (GCP), Docker, AWS Lambda, Azure, iOS, Twilio, Firebase, Linux, MacOS, Kubernetes, Android
Storage
JSON, PostgreSQL, Databases, MongoDB, Redis, Amazon S3 (AWS S3), MySQL, Redis Cache, Microsoft SQL Server
Industry Expertise
Project Management, Healthcare
Other
Debugging, API Integration, APIs, Architecture, Back-end, Back-end Development, Data Scraping, Speech-to-Text (STT), Full-stack, Prompt Engineering, API Design, API Testing, OpenAI GPT-3 API, OpenAI GPT-4 API, Large Language Models (LLMs), Artificial Intelligence (AI), AI Agents, Startups, OpenAI, Webhooks, Full-stack Development, AI Integration, Agentic AI, LangChain, Scripting, Coding, Minimum Viable Product (MVP), SaaS, System Architecture, Retrieval-augmented Generation (RAG), Anthropic, Software Architecture, Software Design, Generative Artificial Intelligence (GenAI), LLM Integration, Text-to-Speech (TTS), Software Development, AI-assisted Development, Monorepos, Bug Fixes, Agentic AI Systems, RAG Architecture, OAuth, Multi-tenant SaaS, Cloud, AI Inference, Optimization, Multithreading, Performance, WebSockets, CI/CD Pipelines, Cloud Architecture, Cloud Storage, FastAPI, Web Applications, Audio Recording, Web Development, Cloud Infrastructure, App Store, Stripe Terminal, User Interface (UI), Agentic RAG Systems, AI Chatbots, Chatbots, Payments, Supabase, Scalability, RAG Systems, Push Notifications, Leadership, API Backwards Compatibility, SSH, AI Agent Orchestration, Systems Engineering, AI Voice Agents, Identity & Access Management (IAM), OpenAI SDK, Agentic Coding, Real Estate, Application Security, AI Engineering, Data Science, Gemini, Front-end, Dashboards, SOC 2, Security, AI Development, AI Evaluation, Solution Architecture, Data Engineering, Concurrency, Scaling, CTO, Cursor AI, Technical Leadership, Linear, Communication, Networking, Machine Learning, JFrog, CloudBees, Sensor Data, GNU, Amazon RDS, Stripe Payments, ElevenLabs Solutions, Figma Make, HIPAA, Amazon Bedrock
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