# Axel Cuevas — Backend & AI Engineer > Backend and AI engineer in Santo Domingo. I'm an AI engineer at Leadsales, building the agents inside a WhatsApp CRM for Latin American businesses. I also run Axiomic, my own company. - Full name: Axel Javier Cuevas Terrero - Location: Santo Domingo, DR · remote - Current roles: AI Engineer at Leadsales; System Architect & Full Stack Developer at Rivero; Freelance Backend Developer & System Architect at IQ BIAI Solutions; Founder & Lead AI Architect at Axiomic LLC; Agentic AI Developer at McTekk SRL - Languages: Spanish · native; English · C2 - Website: https://www.axl-devhub.me - Email: axeljcuevast@gmail.com - LinkedIn: https://www.linkedin.com/in/axl-devhub/ - GitHub: https://github.com/axl-devhub - X: https://x.com/Axl_Caves - CV (PDF): https://www.axl-devhub.me/cv-english.pdf ## Impact ### Leadsales · AI Engineer - 19x lower inference cost for the production Lead Agent in the first week, from a model re-evaluation and migration - Agent replies rewritten to read like a human sales rep instead of generic AI ### Rivero · System Architect & Full Stack Developer - $25k+ saved per pallet-dimensioning station versus off-the-shelf dimensioners, with B2B licensing interest from the DR's largest supermarket chain (RIV-Scan) - ~40% of low-complexity engineering tickets resolved without a human by the Orchy agent - +80% inventory traceability and physical accuracy from the warehouse audit system - ~20% more revenue in under two months from the reporting engine's liquidation and offer views ### IQ BIAI Solutions · Freelance Backend Developer & System Architect - -70% client integration complexity: 3 Dominican payment gateways behind one API - PCI DSS compliance scope reduced to SAQ A through end-to-end tokenization - 100% delivery reliability for payment and refund webhooks ### McTekk SRL · Agentic AI Developer - -60% voice response latency with a real-time audio layer on the Gemini Live API ## Experience ### AI Engineer — Leadsales Sep 2026 – Present · Contract · full-time · Remote Leadsales is a WhatsApp-native CRM and sales platform for LATAM businesses. - Cut the production Lead Agent's inference cost 19x in the first week by re-evaluating model selection and migrating the agent to a far cheaper model. - Rewrote the agent's conversational behavior to strip generic AI phrasing, so replies read like a human sales rep and sell more naturally in customer conversations. - Own the GenAI architecture behind the agents that acquire, qualify and serve Leadsales' customers: RAG over OpenSearch and Pinecone, LLM-based lead classification, multi-format knowledge bases (PDF, Excel, OCR) and human–agent handoff, with cost, latency and quality tracked in production. - Build the team's AI-native workflow: packaged skills, deterministic guardrails around agentic work (hooks, CI gates), documented architecture decisions and mentoring on LLM patterns. Stack: Python, TypeScript, Claude, Pinecone, OpenSearch, LangGraph, MCP ### System Architect & Full Stack Developer — Rivero Oct 2025 – Present · Full-time · Remote · Santo Domingo, DR / Miami, FL Rivero is a logistics and retail group where I own architecture across warehouse, AI and analytics systems. - Engineered a proprietary 3D pallet dimensioning system (hardware and software) using Orbbec ToF depth cameras, WebSockets and a custom Python 3D extraction algorithm, designed to bypass a competitor's US patent, saving $25k+ per station and drawing B2B licensing interest from the DR's largest supermarket chain. - Architected Orchy, a permission-gated AI orchestrator built on OpenClaw and Hermes primitives with strict security scopes across all MCP integrations, so the AI safely executes backend operations (such as Google Workspace provisioning) from Slack. Its self-learning loop resolves 40% of low-complexity engineering tickets while preventing unpermissioned LLM hallucinations. - Owned a mobile-first Warehouse Audit System end to end, replacing blind counts with a risk-based smart routing algorithm and BullMQ for asynchronous pallet re-scoring and audit logging; resolved database-level anomalies (multi-location ghost pallets), raising inventory traceability and accuracy by 80%. - Architected a semantic reporting engine on Cube.js over a PostgreSQL read replica with pre-aggregations, and an interactive AG Grid + ECharts front end with SKU-level drill-downs and actionable views for bulk liquidations and offers, increasing revenue by ~20% in under two months. Stack: TypeScript, Python, NestJS, Next.js, React, PostgreSQL, Redis, BullMQ, ClickHouse, MCP, Docker, Linux ### Freelance Backend Developer & System Architect — IQ BIAI Solutions Jun 2025 – Present · Freelance · Remote Freelance engagement with IQ BIAI Solutions, providing backend development and system architecture services for payment infrastructure. - Architected a multi-tenant payment API in NestJS that unified 3 major Dominican gateways (Cybersource, CardNET, Azul) under a single interface, reducing client integration complexity by 70%. - Engineered a robust security architecture: an HMAC-based API Key system and an end-to-end tokenization flow with Cybersource Flex Microform, reducing PCI DSS compliance scope to SAQ A. - Designed a highly available webhook dispatch system on BullMQ with Redis, ensuring 100% delivery reliability with exponential backoff for critical payment and refund notifications. Stack: NestJS, TypeScript, Node.js, Prisma, Redis, BullMQ, PostgreSQL, Docker ### Founder & Lead AI Architect — Axiomic LLC Jun 2026 – Present · Founder · Remote Axiomic LLC is my own company: DeUna, an autonomous conversational AI sales closer for LATAM SMEs that lives inside WhatsApp. - Founded and architected DeUna, an autonomous conversational AI sales closer for LATAM SMEs, operating natively inside WhatsApp to automate DMs, scheduling, and high-ticket sales. - Engineered a low-latency voice pipeline using Whisper v3 Turbo and ElevenLabs to clone business owners' voices, processing heavy regional dialects for auditory authenticity. - Implemented a bounded-haggling algorithm inside the RAG engine, letting the AI negotiate dynamically with sales psychology while mathematically protecting the hidden price floor. Stack: Python, TypeScript, FastAPI, LangGraph, Pinecone, PostgreSQL, Redis, Docker, AWS ### Agentic AI Developer — McTekk SRL. Sep 2024 – Present · Part-time · Remote McTekk SRL. is a technology company focused on delivering innovative solutions to enhance business operations and customer experiences. - Developed a real-time bidirectional audio layer using WebSockets and the Gemini Live API, reducing voice response latency by 60%. - Engineered robust data pipelines with Dagster, dbt and ClickHouse, applying A/B testing and transformation models. Stack: Python, FastAPI, Gemini, Google ADK, MCP, Dagster, dbt, ClickHouse, Docker ## Skills - Languages: TypeScript, Python, JavaScript, SQL, C#, Go - Backend: NestJS, FastAPI, Node.js, BullMQ, Prisma, Express, ASP.NET Core, WebSockets, GraphQL, AsyncIO, SQLAlchemy, Pydantic, NextAuth, JWT, Slack Bolt, Google Workspace API, Cybersource, CardNET, Azul - Frontend · Mobile: Next.js, React, Tailwind, Vercel, Capacitor, Three.js, Android, AG Grid, ECharts, Bootstrap 5, Framer Motion, ShadCN, Material Tailwind - Data: PostgreSQL, Redis, AWS RDS, ClickHouse, Pinecone, Apache Kafka, MongoDB, Dagster, dbt, Cube.js, Pandas, Jupyter, MySQL, NumPy, Open3D, MariaDB - AI · Agents: Claude Code, LangGraph, MCP, Gemini, Google ADK, Vercel eve, Codex, LangChain, Cursor, Unsloth, OpenClaw, Hermes, A2A protocol - Infra · Cloud: Docker, Git, Linux, GCP, AWS, Railway, Bash, NGINX, Orbbec ToF, Umami ## Education - Pontificia Universidad Católica Madre y Maestra (PUCMM) — B.Sc. in Computer Science Engineering · GPA 4.0 / 4.0 (Jan 2026 — present, Santo Domingo, DR) - Y Combinator Startup School — Summer 2026 cohort (Jul — Aug 2026, San Francisco, CA) - Instituto Técnico Salesiano (ITESA) — Associate's degree in Software Development and Programming · GPA 3.9 / 4 (Aug 2023 — Oct 2025, Santo Domingo, DR) ## Certifications - C2 English Certificate (Band 9.0) · Academia Europea · 2023 - Cybersecurity Fundamentals · LAC4 · 2024 - JavaScript: The Hard Parts · Frontend Masters · 2024 - Complete Intro to Containers · Frontend Masters · 2024 ## Projects ### RIV-Scan — Pallet Dimensioner Hardware · Computer vision · Rivero · 2026 · In production · https://www.axl-devhub.me/en/projects/riv-scan A proprietary 3D pallet dimensioning station, hardware and software, built for Rivero's warehouses. An Orbbec time-of-flight depth camera streams frames over WebSockets into a custom Python 3D extraction pipeline that isolates the pallet from the floor, fits its bounding volume and returns length, width, height and cubic volume to the warehouse system. Designed to bypass a competitor's US patent, it saves over $25k per station versus off-the-shelf dimensioners and drew B2B licensing interest from the country's largest supermarket chain. Problem: Off-the-shelf pallet dimensioners cost more than $25k per station, and the leading vendor sits on a US patent. Rivero's warehouses needed length, width, height and cubic volume for every pallet, fast enough to keep the dock moving. Build: An Orbbec time-of-flight camera streams frames over WebSockets into a Python pipeline that isolates the pallet from the floor, fits its bounding volume and returns the dimensions to the warehouse system. Camera mount, controller and cabling were built in-house around an approach that bypasses the patent. Result: A pallet is measured in under a second, at a fraction of the vendor's price. The design drew B2B licensing interest from the country's largest supermarket chain. Key numbers: $25k+ saved per station; <1 s per pallet; 1 US patent bypassed by design Stack: Python, Open3D, NumPy, Orbbec ToF, WebSockets, FastAPI, Docker ### Orchy — Permission-Gated AI Orchestrator Agents · Orchestration · Rivero · 2026 · Internal · in production · https://www.axl-devhub.me/en/projects/orchy Orchy is Rivero's internal AI orchestrator, built on OpenClaw and Hermes primitives. Every capability is exposed through MCP integrations behind strict security scopes, so the agent executes real backend operations from Slack, such as Google Workspace provisioning or warehouse lookups, without ever acting outside its grant. A self-learning loop turns resolved threads into reusable procedures: it now closes about 40% of low-complexity engineering tickets on its own, while the permission layer prevents unpermissioned LLM hallucinations from reaching production systems. Problem: Routine requests, Google Workspace accounts, warehouse lookups, small fixes, all landed on engineering as Slack threads. An LLM with production access is only acceptable if it cannot act outside what it was granted. Build: Built on OpenClaw and Hermes primitives, every capability is an MCP integration behind an explicit security scope. Orchy reads the thread, checks the grant, executes the backend operation and reports back in Slack; resolved threads feed a self-learning loop that turns them into reusable procedures. Result: About 40% of low-complexity engineering tickets now close without a human. The permission layer keeps unpermitted hallucinations from ever reaching production systems. Key numbers: 40% of low-complexity tickets closed on its own; 100% of capabilities behind a scope; 0 actions outside a grant Stack: Python, OpenClaw, Hermes, MCP, Slack Bolt, Google Workspace API, PostgreSQL ### Rivero Track — Warehouse Platform Warehouse · Platform · Rivero · 2025 — 26 · In production · https://www.axl-devhub.me/en/projects/rivero-track The end-to-end warehouse management platform for Rivero's logistics and retail operation. I helped migrate the backend into a single monorepo and built core features: risk-based cycle-count auditing with smart routing and BullMQ re-scoring, merchandise receiving, the reporting infrastructure, a Cube.js semantic layer over a PostgreSQL read replica with pre-aggregations, an AG Grid and ECharts front end with SKU-level drill-downs, and the integrations that let Orchy and RIV-Scan talk to the same system. Inventory traceability and accuracy went up by 80%; the reporting views drove roughly 20% more revenue in under two months. Problem: Receiving, cycle counts and reporting lived in separate services with drifting data: ghost pallets in several locations at once, blind counts, reports nobody trusted. Build: Migrated the backend into one monorepo and built the core: risk-based cycle-count auditing with smart routing and BullMQ re-scoring, merchandise receiving, and a Cube.js semantic layer over a PostgreSQL read replica feeding an AG Grid and ECharts front end with SKU-level drill-downs. Orchy and RIV-Scan talk to the same system. Result: Inventory traceability and accuracy rose by 80%. The reporting views and actionable liquidation lists drove roughly 20% more revenue in under two months. Key numbers: +80% inventory traceability and accuracy; ~20% more revenue in under two months; 1 monorepo, three products Stack: NestJS, TypeScript, PostgreSQL, BullMQ, Redis, Cube.js, AG Grid, ECharts, Capacitor, Docker ### Multi-Tenant NestJS Payment API Backend · Payments · Freelance · IQ BIAI Solutions · 2025 · In production · https://www.axl-devhub.me/en/projects/payment-api A comprehensive payment processing API built with NestJS that unifies multiple Dominican payment gateways (Cybersource, CardNET, and Azul) under a single interface. Features include complete transaction lifecycle management (Authorization, Capture, Refund, Void), HMAC-based API key authentication, end-to-end tokenization with Cybersource Flex Microform for PCI DSS SAQ A compliance. Problem: Every client taking card payments in the Dominican Republic had to speak three gateway dialects, Cybersource, CardNET and Azul, and carry the full PCI burden of touching card data. Build: A multi-tenant NestJS API with one interface for authorization, capture, refund and void. HMAC-signed API keys, end-to-end tokenization through Cybersource Flex Microform, and a BullMQ/Redis webhook dispatcher with exponential backoff. Result: Client integration complexity dropped by 70%, PCI DSS scope shrank to SAQ A, and payment and refund notifications reach clients with 100% delivery reliability. Key numbers: 3 → 1 gateways behind one interface; -70% integration complexity; SAQ A PCI DSS scope Stack: NestJS, TypeScript, Node.js, Redis, BullMQ, PostgreSQL, Cybersource, CardNET, Azul ### Kanvas AI Multi-agent · Voice · McTekk · 2024 — 25 · Demo · https://www.axl-devhub.me/en/projects/kanvas-ai KanvasAI is a multi-agent system designed to assist business owners by answering natural language questions with precise data information about their business. To achieve that, I collaborated on the creation of a demo that integrated MCP and A2A protocols. It features real-time bi-directional audio communication, allowing users to interact with the system using voice commands. Stack: Google ADK, FastAPI, Gemini, Gemini, AWS, Slack Bolt, GCP ### Aidden - Multi-Agent AI Assistant Multi-agent · Assistant · Personal · 2025 · Demo · https://www.axl-devhub.me/en/projects/aidden A sophisticated voice and text-enabled AI assistant system built with Google's Agent Development Kit (ADK) that orchestrates multiple specialized agents to handle complex, multi-step workflows through natural language interactions. Features intelligent task orchestration with dynamic agent discovery, specialized agents for financial data, filesystem operations, email management, and Obsidian integration. Stack: Google ADK, Python, AsyncIO, MCP, A2A protocol, Natural Language Processing, Gemini, Next.js, Three.js ### Rally Scoring System Real-time · Web app · Personal · 2025 · Live · https://www.axl-devhub.me/en/projects/rally-scoring A real-time scoring management system developed for my school to modernize their rally competitions. The platform allows administrators to manage teams, questions, and stations while enabling real-time score tracking and updates. This system successfully replaced the school's outdated manual scoring method, which was painfully slow as scores were calculated manually after competitions ended. Now, scores are updated automatically and in real-time as each team completes stations, providing immediate feedback and results. Stack: Next.js, Tailwind, Prisma, TypeScript, MongoDB, NextAuth, Vercel, ShadCN Link: https://rally-scoring.vercel.app ### JADAC - Bar Website E-commerce · Personal · 2024 · Archived · https://www.axl-devhub.me/en/projects/jadac A web application allowing users to purchase wines online and redeem promotions at a local bar. Features include browsing wines with tasting notes/pairings/reviews, user accounts for saving favorites and tracking orders. Product information is sourced from a JSON API. Stack: JavaScript, React, Bootstrap 5, Framer Motion, Python, FastAPI, SQLAlchemy, MySQL, Pydantic, Docker, NGINX ### Domu - House Rental Services Rentals · Platform · Personal · 2024 · Archived · https://www.axl-devhub.me/en/projects/domu Developing a modern property rental platform similar to Airbnb. Features include property listing creation/management, advanced booking with calendar integration, real-time availability updates, user authentication/profiles, image uploads, interactive maps, and responsive design. Stack: Next.js, TypeScript, Tailwind, ASP.NET Core, MySQL, JWT, AWS ### Portfolio Web · 3D · Personal · 2024 — 26 · Live · https://www.axl-devhub.me/en/projects/portfolio My personal portfolio website showcasing my projects and skills. It includes a blog section where I share my thoughts on various topics related to software engineering, AI, and Linux. The site is built with Next.js and uses Contentlayer for content management and also contains a 3D model viewer. Stack: Next.js, Tailwind, TypeScript, Prisma, Umami Link: https://axl-devhub.me ### Paycom - Minimalist Payroll System Payroll · Web app · Personal · 2023 · Archived · https://www.axl-devhub.me/en/projects/paycom A user-friendly, web-based application for businesses to manage payroll processes efficiently. Features include employee registration, payment concept modifications, attendance tracking, salary slip viewing, and distinct admin/employee views. Stack: React, Tailwind, Material Tailwind, Python, FastAPI, SQLAlchemy, Pydantic, Docker, MariaDB --- Generated from https://www.axl-devhub.me — the same content as https://www.axl-devhub.me/en/resume and the project pages.