George is available for hire
Hire GeorgeGeorge Bibilashvili
Verified Expert in Engineering
AI and Full-stack Engineer and Developer
Pärnu, Estonia
Toptal member since April 21, 2022
Bio
With over 15 years of experience in software engineering, George is a full-stack AI engineer specializing in AI-driven applications, Python, Agentic AI, TypeScript, React Native, Next.js, React, and Node.js. He's led the development of scalable platforms like UniPay, an eCommerce wallet serving 500,000+ users, and architected production-ready large language model (LLM) integrations that power intelligent RAG architecture, semantic search, and real-world applications.
Portfolio
Healthee
Agentic AI Systems, Agentic RAG Systems, Python, React, React Native, Expo...
Unloop (formerly Sidekick)
React, React Native, AI-augmented Software Engineering (AIASE)...
Tegeta Motors
Agentic AI, Agentic AI Systems, Agentic RAG Systems, AI Integration, SQL...
Experience
- Node.js - 15 years
- TypeScript - 15 years
- Python - 15 years
- Full-stack - 15 years
- FastAPI - 10 years
- React Native - 8 years
- AI Engineering - 5 years
- Agentic AI - 4 years
Work Experience
Full-stack/AI Principal Engineer
2025 - PRESENT
Healthee
- Built and maintained back-end features for Zoe, Healthee's AI health-benefits assistant: a Python/FastAPI service that runs a Pydantic AI agent with 10+ tools over each member's plan.
- Contributed to a latency and cost optimization that fetched the employer-handbook lookup in parallel and merged it into the same agent run.
- Strengthened safety and correctness for a healthcare assistant: input guardrails (crisis detection with a fixed 988/911 response, prompt-injection filtering), wrapping of third-party data as untrusted.
- Helped build the LLM evaluation and release process: 33 routing tests in CI, an eval-mode streaming API used by live-model evals on preview environments for each PR.
Technologies: Agentic AI Systems, Agentic RAG Systems, Python, React, React Native, Expo, AI Agents, Agentic Frameworks, Pydantic, LangChain, LangSmith, Jev, System1, SystemOne, System One, NoSQL, Machine Learning, Swagger, Codex, Databricks, Asyncio, SQLAlchemy, Tailwind UI, Best Practices, Database Migration, Incident Response, Pipelines, Row-level Security (RLS), Role-based Access Control (RBAC), Anthropic, OpenAI API, HIPAA
Lead Full-stack AI Engineer
2024 - 2026
Unloop (formerly Sidekick)
- Led the design and execution of Sidekick’s secure, scalable digital health platform. Defined the technology stack (Node.js, React, and React Native) and deployed it on AWS, integrating with the health API and AI agents.
- Integrated a custom open-source model on Amazon Bedrock, fine-tuned with RAG for our specific domain, and enhanced with custom tools, and provided custom, personalized content for the user.
- Implemented a React Native mobile app and a React web app, handling all communication with AI agents and managing personalized user content.
Technologies: React, React Native, AI-augmented Software Engineering (AIASE), Amazon Bedrock AgentCore, AWS Lambda, AWS CloudFormation, Supabase, PostgreSQL, Node.js, Agentic AI, Model Context Protocol (MCP), Agentic RAG Systems, AI Engineering, Python, FastAPI, Full-stack Development, SaaS, Communication, Code Review, System Architecture, Artificial Intelligence (AI), Software Architecture, CI/CD Pipelines, Unit Testing, E2E Testing, Figma, Claude Code, Cursor AI, API Development, Amazon RDS, Amazon S3 (AWS S3), Django CMS, Django, AI Agents, RESTFul APIs, Test-driven Development (TDD), Git, Vercel, Role-based Access Control (RBAC), Claude API, Claude Agent SDK, Agentic Coding, Claude, Architecture, System Design, RAG Architecture, AI Integration, Data Modeling, OpenAI, Workflow Automation & System Integration, Authentication, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), Google Calendar API, Retrieval-augmented Generation (RAG), System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), AI-assisted Development, Amazon Bedrock, AI Governance, LangGraph, Amazon DynamoDB, Docker, Airtable, AI Tools, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Django REST Framework, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), offline-first local storage and sync (SQLite or WatermelonDB), API Integration, App Store, NoSQL, Machine Learning, Swagger, Codex, Databricks, Asyncio, SQLAlchemy, Tailwind UI, Best Practices, Database Migration, Incident Response, Pipelines, Row-level Security (RLS), Anthropic, OpenAI API, HIPAA
Full-stack AI Principal Engineer
2024 - 2024
Tegeta Motors
- Designed an AI support agent platform where the LLM only proposes typed actions, and a deterministic Policy Engine and Action Gateway decide, enforce human approval, execute, and log every step.
- Critiqued the team's draft design and replaced direct model-to-API calls, 10x retries, and model-sourced refund amounts with a shared gateway, bounded 3-retry idempotent writes, and billing-resolved values.
- Defined a human-in-the-loop tier table enforced in code, not the prompt, keeping refunds, plan changes, and account updates on human approval, with refunds over $500 never auto-approved.
- Engineered a two-pass Action Gateway that resolves server-owned values before policy classification, so prompt injection or a confidently wrong model can never set a monetary value.
- Designed the agent loop with tight budgets of 6 steps, 20s, and a per-case cost cap, where any stop other than resolved-and-replied escalates to a human with the full trace.
- Planned a shadow to assist with the 5% auto-write rollout with a no-deploy kill switch that drops the agent to propose-only on refund-rate, reject-rate, or cost spikes.
- Authored a one-page cross-team agent standard covering tool contracts, idempotency, PII masking, immutable per-case audit records, and required evals and red-team tests before shipping.
Technologies: Agentic AI, Agentic AI Systems, Agentic RAG Systems, AI Integration, SQL, RAG Systems, RAG Architecture, AI Architecture, API Design, eCommerce, Cloudflare Durable Objects, NestJS, SEO Tools, LangChain, GraphRAG, LLM Agents, Knowledge Graphs, Authentication, Relational Databases, Google Calendar API, Retrieval-augmented Generation (RAG), System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, AI-assisted Development, Azure, Amazon Bedrock, AI Governance, LangGraph, Amazon DynamoDB, Docker, Node.js, Architecture, AI Tools, API Development, Continuous Delivery (CD), GitHub Actions, Claude, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), offline-first local storage and sync (SQLite or WatermelonDB), API Integration, App Store, NoSQL, Machine Learning, Swagger, Codex, Databricks, Asyncio, SQLAlchemy, Tailwind UI, Best Practices, Database Migration, Incident Response, Pipelines, Row-level Security (RLS), Role-based Access Control (RBAC), Anthropic, OpenAI API, HIPAA, SOC 2
Full-stack/AI Principal Engineer
2023 - 2024
Telavox
- Designed and built the Django/DRF control plane for an AI phone receptionist, compiling agent prompts and tools into ElevenLabs voice agents and provisioning numbers via Telavox PBX.
- Implemented schema-per-tenant Postgres multi-tenancy with django-tenants, dual tenant resolution, and a read-replica router, eliminating cross-tenant data-leak risk by design.
- Fixed two production Celery incidents on GKE: replaced a starved liveness probe with a heartbeat-file bootstrap, and rescheduled hourly beat jobs to daily to stop ElevenLabs 429s.
- Built an event-driven RAG pipeline on pgvector with HNSW and Vertex AI embeddings (768-dim, EU region), powering semantic call search and a knowledge-gap report of unanswered questions.
- Engineered the latency-sensitive mid-call tool webhook for calendar, CRM, and transfer, masking third-party round trips with pre-tool speech so callers never heard dead air.
- Designed idempotent HMAC-verified post-call webhooks that persist transcripts, fan out audio, indexing, and retention to Celery tasks, and push live events to dashboards via WebSockets.
- Built the integrations layer with Fernet-encrypted credentials, OAuth token refresh jobs, and a CRM/calendar provider registry, and tracked a cleartext-token finding for remediation.
Technologies: Python, React, Amazon Web Services (AWS), ElevenLabs Solutions, Agentic Coding, AI Integration, LLM Integration, AI Tools, AI-assisted Development, Clean Code, Cloudflare, Multi-tenant Architecture, Google Cloud Platform (GCP), Data Security, Authentication, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), Google Calendar API, Retrieval-augmented Generation (RAG), System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, Amazon DynamoDB, Docker, Node.js, Architecture, API Development, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Claude, Django REST Framework, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), offline-first local storage and sync (SQLite or WatermelonDB), API Integration, App Store, NoSQL, Asyncio, SQLAlchemy, Tailwind UI, Best Practices, Database Migration, Incident Response, Pipelines, Row-level Security (RLS), Role-based Access Control (RBAC), OpenAI API
Front-end Developer
2023 - 2024
DesuLabs
- Connected our application to the SUI blockchain using Web3.js, enabling users to interact directly with smart contracts. This involved setting up the necessary infrastructure to handle transactions.
- Implemented functionality to perform various operations on smart contracts, including writing data to the blockchain and retrieving information from it.
- Ensured a stable and user-friendly interface, crucial for the success of our application. Led efforts to maintain and improve the UI, incorporating feedback from users and stakeholders to enhance usability and accessibility.
Technologies: JavaScript, React, Web3.js, Web3, Ethers.js, Games, Gambling, Authentication, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), Google Calendar API, System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, Amazon DynamoDB, Docker, Node.js, Architecture, AI Tools, API Development, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Claude, REST API integration, React Native with Expo (EAS Build and Update), API Integration, App Store, Database Migration, Pipelines, Row-level Security (RLS), Role-based Access Control (RBAC)
Front-end Technical Lead
2022 - 2024
Clearstar
- Contributed to creating an application that gathers valuable data from blockchain smart contracts tailored for traders and investors, elegantly presenting this information across multiple platforms.
- Managed a team of competent front-end developers specializing in React and React Native as a tech lead to execute our development objectives successfully.
- Took an active role in writing code and developing critical components to enhance the scalability and performance of our application, in addition to my leadership responsibilities.
Technologies: React, React Native, JavaScript, Storybook, Leadership, Technical Leadership, Prisma, Amazon Cognito, Amazon Cognito User Pools, AWS Command Line Interface (CLI), AWS Amplify, Node.js, Web3, PostgreSQL, Amazon Web Services (AWS), Scrum, Minimum Viable Product (MVP), Full-stack, Python 3, FastAPI, Authentication, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, Amazon DynamoDB, Docker, Architecture, AI Tools, API Development, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Claude, Django REST Framework, REST API integration, camera and photo capture, API Integration, App Store, Pipelines, Role-based Access Control (RBAC)
Mobile Developer
2023 - 2023
Toothy LLC
- Facilitated the process of eliciting comprehensive client requirements effectively, culminating in the selection of an optimal technology stack. Used React Native and Firebase.
- Orchestrated the architectural design of both the back-end infrastructure and the mobile application, leveraging the React Native framework. Executed the integration of a bespoke chatbot tailored for mobile users.
- Prepared an application for internal testing and released it internally for local users.
Technologies: Chatbots, React Native, iOS, Android, App Store, App Store Acceptance Process, Firebase, 3D Graphics, Mobile Back-end, Google Play Store, Prisma, Amazon Cognito, Amazon Cognito User Pools, Flask, Search Engine Optimization (SEO), Agile, Scrum, Mobile Development, Minimum Viable Product (MVP), Full-stack, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, Docker, Node.js, Architecture, Medplum, Healthtech, AI Tools, API Development, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), camera and photo capture, API Integration, Pipelines
Senior Full-stack Web3.js Developer
2023 - 2023
Aragon
- Collaborated with skilled blockchain developers to create a groundbreaking Dapp enabling users to establish DAOs via a Metamask-linked web app, with features such as proposal creation, token allocation, naming, and a robust voting system.
- Structured a project as Monorepo with three components: React provided a smooth user experience, and the design system was built with Storybook and testing. The SDK efficiently communicated with smart contracts governing our DAO's logic.
- Developed a web app and aided in resolving any issues that arose, including front-end bugs and smart contract logic flows. Implemented updates to the SDK, ensuring smooth, long-term support for the optimal functionality of our product.
Technologies: Next.js, React, Solidity, Decentralized Autonomous Organizations (DAO), Ethereum, JavaScript, ERC-20, MetaMask, Infura, IPFS, Apollo, GraphQL, Web3.js, DApps, Web App Development, HTML5, CSS, Storybook, Amazon Cognito, Amazon Cognito User Pools, Stripe, Web3, Amazon Web Services (AWS), Search Engine Optimization (SEO), Agile, Scrum, Minimum Viable Product (MVP), Agile Project Management, Relational Databases, System Integration, Data Pipelines, Service-oriented Architecture (SOA), System Design, Docker, Node.js, Architecture, AI Tools, Continuous Delivery (CD), GitHub Actions, REST API integration, React Native with Expo (EAS Build and Update), camera and photo capture, API Integration, App Store, Pipelines
Senior React Engineer
2022 - 2023
Tinloof GmbH
- Achieved a 100% score on Google speed tools by building top-quality user interfaces focused on high performance. Created eCommerce, fintech, and social applications with the new, fine-tuned UI.
- Developed the web portal Goalimpact, a database of football players and their statistics with the ability to subscribe to each player as a fan or scout and manage data.
- Empowered retailers to effortlessly generate advertising revenue with the Retailmediatools web platform I helped develop. Served as the senior front-end force on the team.
- Created a progressive web application, Match Cards, utilizing React, TypeScript, and Next.js. Designed to enhance memory retention through playful emoji cards, it offers a fully-immersive, native-like experience on mobile devices.
- Participated in creating the marketplace from scratch for an international gallery that displays limited edition contemporary pieces of art in furniture.
- Established a branding website by creating a schema and architecture of the framework with React, Next.js, TypeScript, and Sanity as a Headless CMS. Achieved a fully dynamic content management system and high-performance UI.
- Organized content writing. Wrote an interesting article titled, "What is PWA?" This outlined the benefits of advanced web applications and how to win the mobile market at a low cost easily.
Technologies: TypeScript, HTML, JavaScript, CSS, React, Web Development, Next.js, Gatsby, Remix IDE, Headless Software, API Integration, CTO, Leads, Sanity Testing, Firebase, Full-stack, GraphQL, Node.js, REST, Web App Development, APIs, Redux, Redux-Saga, HTML5, Express.js, Mobile, Mobile App Development, Bootstrap, User Interface (UI), Front-end, Databases, WebSockets, React Hooks, Jest, Styled-components, Progressive Web Applications (PWAs), PostgreSQL, Amazon Web Services (AWS), Animatron, 3D Games, CSS3, GitHub, Tailwind CSS, Supabase, Google Analytics, Team Leadership, UX Design, User Experience (UX), Back-end, Python, SQL, Firebase Authentication, JSON, Agile Software Development, DOM, MongoDB, Material UI, Chakra UI, Remix.run, Sanity Studio, Storybook, PostCSS, Emotion Library, ECMAScript (ES6), Serverless, Prisma, Amazon Cognito, Amazon Cognito User Pools, Stripe, Flask, Zustand, Search Engine Optimization (SEO), Agile, Scrum, Minimum Viable Product (MVP), Agile Project Management, Relational Databases, System Integration, Data Pipelines, Service-oriented Architecture (SOA), System Design, Docker, Architecture, GitHub Actions, REST API integration, React Native with Expo (EAS Build and Update), App Store, Shadcn UI
React Native Senior Mobile Developer
2013 - 2023
Happy Ventures Public Benefit Corporation
- Developed a high-performance mobile application using React Native, increasing user engagement by 30% within the 1st three months. Integrated cross-platform features for both iOS and Android.
- Tracked and resolved over 200 bugs and performance issues using Jira as a reporting tool. Led sprint retrospectives to improve issue-tracking efficiency by 25%, resulting in faster bug resolution times and higher app stability.
- Led the development of a cross-platform mobile application using React Native, ensuring seamless compatibility across both iOS and Android platforms.
- Spearheaded the launch of a React Native application, overseeing the end-to-end deployment process on both iOS and Android platforms.
Technologies: React Native, HTML, CSS, JavaScript, Expo, Google Play Store, App Store, Firebase, Push Notifications, React Query, Atom, Biometrics, Security, Cursor AI, Claude, AWS Lambda, AWS CloudFormation, Supabase, PostgreSQL, Node.js, Agentic AI, Model Context Protocol (MCP), Agentic RAG Systems, AI Engineering, Python, FastAPI, Full-stack Development, SaaS, Communication, Code Review, System Architecture, Artificial Intelligence (AI), Software Architecture, CI/CD Pipelines, Unit Testing, E2E Testing, Figma, Test-driven Development (TDD), Claude Code, API Development, Amazon RDS, Amazon S3 (AWS S3), Django, Django CMS, AI Agents, RESTFul APIs, Git, Vercel, Claude API, Claude Agent SDK, Authentication, Relational Databases, Fast Healthcare Interoperability Resources (FHIR), Google Calendar API, System Integration, Data Pipelines, HIPAA Compliance, Service-oriented Architecture (SOA), System Design, AI-assisted Development, Amazon DynamoDB, Docker, Architecture, AI Tools, Continuous Delivery (CD), Continuous Integration (CI), GitHub Actions, Django REST Framework, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), offline-first local storage and sync (SQLite or WatermelonDB), API Integration
Senior React Native Developer
2022 - 2022
Blapp Inc
- Developed an app to find black-owned businesses in the users' neighborhood for the USA market.
- Launched the application after completing the build of essential features. Next, we added features making it more social with the ability to create profiles. This resulted in a streamlined onboarding flow and the ability to follow users.
- Improved the app's search function of people with added verification steps, user location, and phone verification. Used Twilio for sending SMS'.
- Implemented live data using Firebase real-time systems with Firestore. Features included weekly blappers, most blapped in the week, and other live statistical data on the application dashboard to make it more dynamic.
Technologies: React Native, Firebase, Expo.io, TypeScript, Firebase Cloud Functions, Firebase Realtime Database, Authorization, iOS, Android, JavaScript, React, Jest, Styled-components, Animatron, CSS3, GitHub, Team Leadership, UX Design, User Experience (UX), Firebase Authentication, JSON, Agile Software Development, DOM, REST, APIs, Next.js, Responsive Web Design (RWD), Firebase Cloud Messaging (FCM), Mobile App Design, App Design, ECMAScript (ES6), Stripe, Flask, Amazon Web Services (AWS), Search Engine Optimization (SEO), Agile, Scrum, Mobile Development, Minimum Viable Product (MVP), Agile Project Management, Full-stack, Authentication, Relational Databases, System Integration, Data Pipelines, Service-oriented Architecture (SOA), System Design, Docker, Architecture, REST API integration, API Integration, App Store
Senior React Native Developer
2020 - 2022
StartupCraft
- Developed mobile apps for clients using React Native, TypeScript, GraphQL, Redux, and React-Query with a team of passionate React Native developers.
- Founded OhanaLink Technologies with a mission to deliver solutions to preserve privacy, intimacy, and availability for groups of close people such as family, friends, or communities.
- Grew WorkAxle, a one-stop solution for chaotic workplaces, to 30,000 online users, transformed it into an enterprise solution, and raised CAD 2.8 million. This solution helps manage jobs through schedules.
Technologies: React Native, JavaScript, TypeScript, React, GraphQL, Mobile, Mobile App Development, Android, iOS, Web3.js, Redux, Redux-Saga, Redux Thunk, Jest, Animatron, CSS3, GitHub, UX Design, User Experience (UX), Firebase Authentication, JSON, Agile Software Development, DOM, REST, APIs, Responsive Web Design (RWD), Mobile App Design, App Design, App Development, ECMAScript (ES6), Leadership, Flask, Zustand, Search Engine Optimization (SEO), Agile, Scrum, Mobile Development, Minimum Viable Product (MVP), Agile Project Management, Full-stack, Authentication, Relational Databases, System Integration, Data Pipelines, Service-oriented Architecture (SOA), System Design, Architecture, REST API integration, API Integration, App Store
Senior Full-stack Developer
2017 - 2020
UniPAY
- Developed a merchant payment gateway system for an eCommerce website. It is a Stripe-like solution for the Georgian market.
- Created a service payment gateway with Node.js as the main REST API and implemented approximately 50 local payment services.
- Built the Georgian Railway online ticket market enabling customers to buy metro tickets online.
- Created IOS and Android mobile applications using React Native connected to our service gateway. With this integration, users can pay for different services and goods from our mobile app.
- Designed a crypto wallet to manage crypto assets such as Bitcoin, ETH, DOGE etc. Users can sell, buy, and transfer crypto money easily.
Technologies: JavaScript, TypeScript, React, Full-stack, Node.js, REST, Bootstrap, HTML, HTML5, API Integration, MySQL, Progressive Web Applications (PWAs), Firebase, Web Development, Front-end, User Interface (UI), CSS, Databases, REST APIs, Express.js, MongoDB, WebSockets, Next.js, Vue, React Native, iOS, Android, Mobile, CSS3, GitHub, Back-end, Python, SQL, JSON, AWS Amplify, DOM, Redis, APIs, PostgreSQL, Material UI, Chakra UI, Responsive Web Design (RWD), Heroku, Storybook, Docker, PostCSS, App Design, App Development, Serverless, Leadership, Prisma, Amazon Cognito, Amazon Web Services (AWS), Zustand, Search Engine Optimization (SEO), Agile, Scrum, Minimum Viable Product (MVP), Agile Project Management, Authentication, Relational Databases, Service-oriented Architecture (SOA), System Design, App Store
React Native Developer
2015 - 2017
Bridge
- Built pixel-perfect, buttery-smooth UIs across two mobile platforms.
- Leveraged native APIs for deep integrations with both platforms. Diagnosed and fixed bugs and performance bottlenecks for performance that feels native.
- Reached out to the open-source community to encourage and assist in implementing mission-critical software fixes.
Technologies: React Native, JavaScript, TypeScript, React, GraphQL, Android, iOS, Redux, Redux-Saga, Redux Thunk, HTML5, Mobile, User Interface (UI), Front-end, React Hooks, Mobile App Development, Progressive Web Applications (PWAs), Jest, CSS3, GitHub, SQL, JSON, DOM, REST, APIs, Node.js, PostgreSQL, MongoDB, Next.js, Material UI, Responsive Web Design (RWD), Mobile App Design, Storybook, App Development, Amazon Web Services (AWS), Agile, Mobile Development, Minimum Viable Product (MVP), Full-stack, Docker, App Store
Project History
Hybrid Search RAG: Cited Answers from 57k Finance Posts
https://rag.lynxly.ai/A chat assistant that answers personal-finance questions from 57,638 real forum posts (the FiQA dataset) and cites the post behind every claim. I designed and built it end to end as a solo project.
Search runs entirely in memory, with no vector database. BM25 keyword search and OpenAI embeddings each search both the user's question and the agent's rewritten query. Reciprocal Rank Fusion merges the four result lists, and a Cohere reranker picks the top 8. A Pydantic AI agent answers only from those posts. A FastAPI streaming API sends the answer to a React frontend that shows what each search stage returned, live.
I built an evaluation suite to measure quality. For 92% of test questions, the correct post is in the top 8 results. An LLM judge checks every claim against its cited source, and 99% are backed by it. The repo also includes a script and guide for indexing your own documents, and it deploys to Railway with Docker.
Search runs entirely in memory, with no vector database. BM25 keyword search and OpenAI embeddings each search both the user's question and the agent's rewritten query. Reciprocal Rank Fusion merges the four result lists, and a Cohere reranker picks the top 8. A Pydantic AI agent answers only from those posts. A FastAPI streaming API sends the answer to a React frontend that shows what each search stage returned, live.
I built an evaluation suite to measure quality. For 92% of test questions, the correct post is in the top 8 results. An LLM judge checks every claim against its cited source, and 99% are backed by it. The repo also includes a script and guide for indexing your own documents, and it deploys to Railway with Docker.
Zoe – AI Health Benefits Assistant (Healthee)
https://healthee.comZoe is Healthee's AI virtual assistant that helps employees understand their health benefits: deductibles and out-of-pocket spend, what's covered and what it will cost, in-network providers, prescriptions, and employer-sponsored programs. As a Backend/AI Engineer, I worked on the Python service behind Zoe. It authenticates the member, gathers their plan and benefits data from internal services in parallel, runs a tool-calling LLM agent (e.g., Pydantic AI or OpenAI) with 10+ tools, and streams answers in real time via GraphQL WebSocket subscriptions. I worked on agent tools and prompt templates; guardrails for a healthcare setting (crisis detection, prompt-injection defense, grounding checks so the assistant never states numbers without plan data); and LLM evaluation, from CI routing tests to live model evals in per-PR preview environments. I also contributed to a performance change that cut response latency by about 26% with no drop in answer quality, measured by a blind A/B test. The service runs on AWS EKS with Postgres, OpenTelemetry tracing, Prometheus, and Sentry.
SEC 10-K GraphRAG Assistant (LangGraph, Neo4j, LangSmith)
An AI research assistant that answers natural-language questions about the latest annual reports (SEC 10-K filings) of 22 US semiconductor companies, such as NVIDIA, AMD, Intel, and TSMC's main customers, and cites a source for every claim. I designed and built it end-to-end in Python. It downloads filings from the SEC's EDGAR database, splits them into token-sized chunks, and loads them into Neo4j with embeddings and a vector index. Re-running the pipeline only re-embeds text that changed. On top of that, I hand-built a LangGraph tool-calling agent that decides when to search, can search per company for comparisons, maintains conversation memory, and streams answers in a CLI chat. Every step is traced in LangSmith. After the model once requested 46 searches in a single reply, I added a hard search budget in code that forces an answer from the results already found. In progress: building a knowledge graph of suppliers, customers, and risks, a Text2Cypher tool, and LangSmith experiments that measure where graph retrieval beats plain vector RAG.
Agent Loop | RAG for Unloop Journaling and Healthee Analysis
Led the design and execution of Sidekick’s secure, scalable digital health platform. Defined the technology stack (Node.js, React, and React Native) and deployed it on AWS, integrating with the health API and AI agents.
Integrated a custom open-source model on Amazon Bedrock, fine-tuned with RAG for our specific domain, and enhanced with custom tools. Provided custom, personalized content for the user.
Implemented a React Native mobile app and React web app, handling all communication with AI agents and managing personalized user content.
Integrated a custom open-source model on Amazon Bedrock, fine-tuned with RAG for our specific domain, and enhanced with custom tools. Provided custom, personalized content for the user.
Implemented a React Native mobile app and React web app, handling all communication with AI agents and managing personalized user content.
Customer Service AI Agent + Knowledge Base
Designed an AI support agent platform where the LLM only proposes typed actions, and a deterministic Policy Engine and Action Gateway decide, enforce human approval, execute, and log every step.
Critiqued the team's draft design and replaced direct model-to-API calls, 10x retries, and model-sourced refund amounts with a shared gateway, bounded 3-retry idempotent writes, and billing-resolved values.
Defined a human-in-the-loop tier table enforced in code, not the prompt, keeping refunds, plan changes, and account updates on human approval, with refunds over $500 never auto-approved.
Engineered a two-pass Action Gateway that resolves server-owned values before policy classification, so prompt injection or a confidently wrong model can never set a monetary value.
Designed the agent loop with tight budgets of 6 steps, 20s, and a per-case cost cap, where any stop other than resolved-and-replied escalates to a human with the full trace.
Planned a shadow-to-assisted 5% auto-write rollout with a no-deploy kill switch that drops the agent to propose-only on refund rate, reject rate, or cost spikes.
Critiqued the team's draft design and replaced direct model-to-API calls, 10x retries, and model-sourced refund amounts with a shared gateway, bounded 3-retry idempotent writes, and billing-resolved values.
Defined a human-in-the-loop tier table enforced in code, not the prompt, keeping refunds, plan changes, and account updates on human approval, with refunds over $500 never auto-approved.
Engineered a two-pass Action Gateway that resolves server-owned values before policy classification, so prompt injection or a confidently wrong model can never set a monetary value.
Designed the agent loop with tight budgets of 6 steps, 20s, and a per-case cost cap, where any stop other than resolved-and-replied escalates to a human with the full trace.
Planned a shadow-to-assisted 5% auto-write rollout with a no-deploy kill switch that drops the agent to propose-only on refund rate, reject rate, or cost spikes.
AI Phone Receptionist Platform (Telavox × ElevenLabs)
http://www.telavox.comDesigned and built the Django/DRF control plane for an AI phone receptionist, compiling agent prompts and tools into ElevenLabs voice agents and provisioning numbers via Telavox PBX.
Implemented schema-per-tenant Postgres multi-tenancy with django-tenants, dual tenant resolution, and a read-replica router, eliminating cross-tenant data-leak risk by design.
Fixed two production Celery incidents on GKE: replaced a starved liveness probe with a heartbeat-file bootstep, and rescheduled hourly beat jobs to daily to stop ElevenLabs 429s.
Built an event-driven RAG pipeline on pgvector with HNSW and Vertex AI embeddings (768-dim, EU region), powering semantic call search and a knowledge-gap report of unanswered questions.
Engineered the latency-sensitive mid-call tool webhook for calendar, CRM, and transfer, masking third-party round trips with pre-tool speech so callers never heard dead air.
Designed idempotent HMAC-verified post-call webhooks that persist transcripts, fan out audio, indexing, and retention to Celery tasks, and push live events to dashboards via WebSockets.
Implemented schema-per-tenant Postgres multi-tenancy with django-tenants, dual tenant resolution, and a read-replica router, eliminating cross-tenant data-leak risk by design.
Fixed two production Celery incidents on GKE: replaced a starved liveness probe with a heartbeat-file bootstep, and rescheduled hourly beat jobs to daily to stop ElevenLabs 429s.
Built an event-driven RAG pipeline on pgvector with HNSW and Vertex AI embeddings (768-dim, EU region), powering semantic call search and a knowledge-gap report of unanswered questions.
Engineered the latency-sensitive mid-call tool webhook for calendar, CRM, and transfer, masking third-party round trips with pre-tool speech so callers never heard dead air.
Designed idempotent HMAC-verified post-call webhooks that persist transcripts, fan out audio, indexing, and retention to Celery tasks, and push live events to dashboards via WebSockets.
Didit.me
I developed DApp applications using React and React Native for Gamium's Blockchain Identity Platform. I implemented a structured monorepo setup for both applications, enhancing code organization and sharing of common libraries. Also, I established a robust state management system for seamless application flow and data handling. Further, the Wagmi library and WalletConnect were utilized for secure communication with the blockchain network. Additionally, I designed a user-friendly platform allowing individuals to connect to the blockchain, input personal information, and establish a secure identity.
Build Your DAO (Smart Contract DApp Web3.js)
Build your DAO, explore communities, and find inspiration for your project all in one place. Mint tokens, set governance parameters, and deploy your DAO on-chain in minutes with our no-code set up process.
I collaborated with skilled blockchain developers to create a groundbreaking Dapp enabling users to establish DAOs via a Metamask-linked web app, with features such as proposal creation, token allocation, naming, and a robust voting system. I structured a project as a monorepo with three components; React provided a smooth user experience, and the design system was built with Storybook and testing. The SDK efficiently communicated with smart contracts governing our DAO's logic.
Additionally, I developed a web app and aided in resolving any issues that arose, including front-end bugs and smart contract logic flows. Implemented updates to the SDK, ensuring smooth long-term support for the optimal functionality of our product.
I collaborated with skilled blockchain developers to create a groundbreaking Dapp enabling users to establish DAOs via a Metamask-linked web app, with features such as proposal creation, token allocation, naming, and a robust voting system. I structured a project as a monorepo with three components; React provided a smooth user experience, and the design system was built with Storybook and testing. The SDK efficiently communicated with smart contracts governing our DAO's logic.
Additionally, I developed a web app and aided in resolving any issues that arose, including front-end bugs and smart contract logic flows. Implemented updates to the SDK, ensuring smooth long-term support for the optimal functionality of our product.
ClearSTAR Trading Platform (Invite Only)
I played a pivotal role in the collaborative development of an application, specifically focusing on extracting valuable insights from blockchain smart contracts. Our target audience encompassed both traders and investors, and our aim was to deliver this information in an elegant and user-friendly manner, ensuring accessibility across various platforms.
In my capacity as a tech lead, I had the privilege of skillfully managing a proficient team of front-end developers who specialized in React and React Native. This leadership role was instrumental in achieving our development goals with remarkable success. I assumed an active, hands-on role, not only in guiding the team but also in writing code and crafting critical components that significantly contributed to the scalability and performance enhancement of our application. This multifaceted approach was crucial in driving our project towards its successful execution.
In my capacity as a tech lead, I had the privilege of skillfully managing a proficient team of front-end developers who specialized in React and React Native. This leadership role was instrumental in achieving our development goals with remarkable success. I assumed an active, hands-on role, not only in guiding the team but also in writing code and crafting critical components that significantly contributed to the scalability and performance enhancement of our application. This multifaceted approach was crucial in driving our project towards its successful execution.
Benable
https://benable.com/I led the development of a cross-platform mobile application using React Native, ensuring seamless compatibility across both iOS and Android platforms. Further, I reduced overall development time by 50% by leveraging reusable components and efficient state management solutions such as MobX. I also integrated 3rd-party services like Firebase for real-time data synchronization, which improved user engagement by 30% within three months of launch. I also applied performance optimization techniques, including lazy loading and code-splitting, which decreased the app's initial load time by 40%. The project's scalable architecture allowed the app to support new feature rollouts without significant refactoring. By following best practices and maintaining clean, modular code, I ensured the app was easily maintainable for future iterations.
HumanOS
https://humanos.co.uk/As a React/React Native developer immersed in the dynamic world of fitness technology, I led the development of a transformative web/mobile application to revolutionize personal wellness journeys. From conception to realization, I orchestrated the creation of an intuitive platform tailored for the fitness industry, enabling users to maintain their shape, track habits and goals effortlessly, and seamlessly communicate with coaches for personalized recommendations.
Our application leveraged the power of React and React Native to deliver a seamless user experience across iOS and Android devices, ensuring accessibility for all fitness enthusiasts. By integrating cutting-edge features such as habit tracking, goal setting, and real-time chat functionality with coaches, we empowered users to take charge of their fitness journeys like never before.
Our solution elevated the standard of fitness applications and fostered a supportive community where users could thrive, achieve their goals, and embark on transformative wellness journeys with confidence.
Our application leveraged the power of React and React Native to deliver a seamless user experience across iOS and Android devices, ensuring accessibility for all fitness enthusiasts. By integrating cutting-edge features such as habit tracking, goal setting, and real-time chat functionality with coaches, we empowered users to take charge of their fitness journeys like never before.
Our solution elevated the standard of fitness applications and fostered a supportive community where users could thrive, achieve their goals, and embark on transformative wellness journeys with confidence.
Blapp | Black-owned Businesses
https://apps.apple.com/us/app/blapp-find-black-owned/id1573799218Blapp makes it easy for you to find and support black-owned businesses across the US, where the tech stack includes React Native, TypeScript, Firebase Auth, Firebase Database, and Firebase Cloud functions. We added more features to the application. I made the application more social and added new functionality, including creating profiles, onboarding flow, and the ability to follow users and get followers. Improved searching for other people in the application with some verification steps to add a location for the user and phone verification. We used Twilio service for sending SMS functionality. Implemented live data using Firebase real-time systems with Firestore. Features like Weekly Blappers, Most Blapped in the week, and other live statistical data on the application dashboard to make it more dynamic.
OhanaLink Technologies | Private Communication
OhanaLink Technologies is on a mission to deliver solutions that preserve privacy, intimacy, and availability for groups of close people such as family, friends, or communities. It allows users to own their small Instagram-like news feed, chat, and custom content fully protected from the outer world.
OhanaLink requirements ran real-time communication, such as chat, a news feed, and other features. We faced the challenge of delivering a cost-efficient and attractive solution simultaneously. We did this with React Native and GraphQL. The application was made to be a fully native experience.
It's impossible to imagine communication without online chat these days. Customers need a 1-1 texting tool and the ability to send media content such as audio and video, have likes and emojis, and have group conversations. Having complete control over who has access and where was significantly important. The back-end team used React and GraphQL to build a fantastic chat experience on the React Native app. I implemented one of the most popular features known these days, an Instagram-like news feed.
OhanaLink requirements ran real-time communication, such as chat, a news feed, and other features. We faced the challenge of delivering a cost-efficient and attractive solution simultaneously. We did this with React Native and GraphQL. The application was made to be a fully native experience.
It's impossible to imagine communication without online chat these days. Customers need a 1-1 texting tool and the ability to send media content such as audio and video, have likes and emojis, and have group conversations. Having complete control over who has access and where was significantly important. The back-end team used React and GraphQL to build a fantastic chat experience on the React Native app. I implemented one of the most popular features known these days, an Instagram-like news feed.
UniPay Payment System
https://unipay.com/UniPay offers a universal payment system in Georgia, handling over 100,000 users and 100 transactions per second. My team and I built and optimized this application with PHP and Node.js microservices. We used React, Next.js, and styled components with MySQL and MongoDB on the back-end side.
Bridge App
https://beta.brdg.app/Connect with people easily, yet personally. Offer intros at scale, discover warm introductions, uncover insights, and connect the dots in real-time with the people you trust. Save time with personalized templates, automatically saved snippets, and tracking. No more tracking email threads and follow-ups.
WorkAxle | Work Scheduler
https://apps.apple.com/us/app/workaxle/id1313407282?ls=1The WorkAxle is a one-stop solution against chaotic workplaces where the job is managed through schedules. Automatic shift scheduling was our top priority for the project. By utilizing data such as worker availability, vacations, holidays, time-offs, and pre-configured rules, we left almost no work for managers.
Clock-in with beacons and face recognition makes this product unique, along with the number of bleeding-edge technologies we use. The clock-in and clock-out with automatic time and location tracking are empowered by face recognition, automated time and attendance rules, and real-time notifications, interconnected altogether. As a team of React Native gurus, we created unique solutions for cross-platform apps using React, React Native, graphQL, Redux, Redux-saga, and TypeScript.
Clock-in with beacons and face recognition makes this product unique, along with the number of bleeding-edge technologies we use. The clock-in and clock-out with automatic time and location tracking are empowered by face recognition, automated time and attendance rules, and real-time notifications, interconnected altogether. As a team of React Native gurus, we created unique solutions for cross-platform apps using React, React Native, graphQL, Redux, Redux-saga, and TypeScript.
AI Assistant & Training Platform for Restaurant Staff
https://ostati.ai/en/Ostati.ai is a training platform for restaurants in Georgia that helps waiters learn service skills through short lessons, 3D practice, and an AI assistant they can ask during a shift.
I built the AI support agent end-to-end: the full agentic loop and the knowledge base behind it. Staff ask questions by chat or voice in Georgian, English, or Russian, for example, how to handle a guest's allergy or a problem at the table, and the agent searches the knowledge base, decides which tools to use, and answers in the staff member's language.
The knowledge base combines the platform's training content with each venue's own manager tips, and the agent checks the manager's tips first, so answers follow how that specific restaurant works instead of generic advice. I designed the retrieval, the agent loop with its tools and stopping rules, and the grounding so answers come only from the knowledge base.
I built the AI support agent end-to-end: the full agentic loop and the knowledge base behind it. Staff ask questions by chat or voice in Georgian, English, or Russian, for example, how to handle a guest's allergy or a problem at the table, and the agent searches the knowledge base, decides which tools to use, and answers in the staff member's language.
The knowledge base combines the platform's training content with each venue's own manager tips, and the agent checks the manager's tips first, so answers follow how that specific restaurant works instead of generic advice. I designed the retrieval, the agent loop with its tools and stopping rules, and the grounding so answers come only from the knowledge base.
Education
2007 - 2011
Bachelor's Degree in Computer Science
The University of Georgia - Tbilisi, Georgia
Skills
Libraries/APIs
React, Node.js, Redux-Saga, REST APIs, React Router, Twilio API, JSONP, Emotion Library, Stripe, API Development, Google Play API, Google Calendar API, Asyncio, SQLAlchemy, OpenAI API, Web3.js, AWS Amplify, Vue, Wagmi, React Query, Claude API, Pydantic, OpenAPI
Tools
Redux Thunk, Expo.io, Firebase Cloud Messaging (FCM), Remix IDE, Auth0, Bitbucket, GitHub, Google Analytics, Firebase Authentication, MetaMask, Prisma, Amazon Cognito, AWS Command Line Interface (CLI), Expo, Claude, Claude Code, Codex, Shadcn UI, Animatron, PostCSS, Atom, AWS CloudFormation, Figma, Git, Claude Agent SDK, GraphRAG, Okapi BM25
Languages
JavaScript, TypeScript, GraphQL, HTML, CSS, HTML5, CSS3, Python, SQL, ECMAScript (ES6), Python 3, Solidity, Cypher
Frameworks
React Native, Next.js, Redux, Express.js, Bootstrap, Jest, Tailwind CSS, Flask, LangGraph, Swagger, Material UI, Django REST Framework, Django, NestJS, Agentic Frameworks
Paradigms
REST, UX Design, Agile Software Development, Responsive Web Design (RWD), Mobile App Design, App Development, Mobile Development, Search Engine Optimization (SEO), Agile, Scrum, Agile Project Management, Model Context Protocol (MCP), Role-based Access Control (RBAC), Fast Healthcare Interoperability Resources (FHIR), HIPAA Compliance, Service-oriented Architecture (SOA), Microservices, Continuous Delivery (CD), Continuous Integration (CI), Sanity Testing, Best Practices, Unit Testing, E2E Testing, Test-driven Development (TDD), Clean Code
Platforms
Android, iOS, Firebase, Mobile, Bitrise, Heroku, Sanity Studio, Docker, Ethereum, Blockchain, Vercel, Amazon Web Services (AWS), Twilio, Medplum, AWS Lambda, RevenueCat, Django CMS, Google Cloud Platform (GCP), Azure, LangSmith, Langfuse, Databricks
Storage
PostgreSQL, MySQL, Databases, Firebase Realtime Database, Cloud Firestore, JSON, Relational Databases, Data Pipelines, Amazon DynamoDB, NoSQL, Database Migration, MongoDB, Redis, Amazon S3 (AWS S3), Neo4j
Other
Full-stack, Web Development, APIs, API Integration, Mobile App Development, User Interface (UI), Front-end, WebSockets, React Hooks, Firebase Cloud Functions, Authorization, Front-end Development, Firebase Hosting, Web App Development, TypeORM, Styled-components, Headless Software, CTO, Leads, PWA, Progressive Web Applications (PWAs), Supabase, Team Leadership, User Experience (UX), Back-end, DOM, Chakra UI, Remix.run, Storybook, App Design, Serverless, Decentralized Autonomous Organizations (DAO), Infura, IPFS, DApps, Smart Contracts, Ethereum Smart Contracts, ERC-20, ERC-721, Ethers.js, Decentralized Applications, Leadership, Technical Leadership, Chatbots, App Store, App Store Acceptance Process, 3D Graphics, Mobile Back-end, Google Play Store, Web3, Amazon Cognito User Pools, Zustand, Minimum Viable Product (MVP), Agentic AI, FastAPI, Full-stack Development, SaaS, Artificial Intelligence (AI), Software Architecture, CI/CD Pipelines, AI Agents, Architecture, System Design, Large Language Models (LLMs), AI Integration, AI Tools, AI-assisted Development, LangChain, Authentication, OAuth, Calendars, Retrieval-augmented Generation (RAG), System Integration, Amazon Bedrock, AI Governance, GitHub Actions, App Store and Google Play releases, Offline Data Synchronization, REST API integration, React Native with Expo (EAS Build and Update), camera and photo capture, offline-first local storage and sync (SQLite or WatermelonDB), signature capture, Software Engineering, Tailwind UI, Pipelines, Anthropic, Gatsby, 3D Games, Apollo, Healthtech, Airtable, Incident Response, Row-level Security (RLS), HIPAA, Cryptocurrency Wallets, WalletConnect, Games, Gambling, Push Notifications, Biometrics, Security, Cursor AI, AI-augmented Software Engineering (AIASE), Amazon Bedrock AgentCore, Agentic RAG Systems, AI Engineering, Communication, Code Review, System Architecture, Amazon RDS, RESTFul APIs, Agentic Coding, RAG Architecture, Monitoring, Solution Architecture, Enterprise Architecture, LLM Integration, Agentic AI Systems, Data Modeling, OpenAI, Workflow Automation & System Integration, ElevenLabs Solutions, Cloudflare, RAG Systems, AI Architecture, API Design, eCommerce, Cloudflare Durable Objects, SEO Tools, Multi-tenant Architecture, Data Security, LLM Agents, Knowledge Graphs, Workflow Automation, Systems Design, AI Infrastructure, Multi-agent Orchestration, Jev, System1, SystemOne, System One, Machine Learning, SOC 2, AI Development, AI Chatbots, Coherent UI, Hybrid Search, LLM Evaluation, Railway
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