Benjamin Berta · B2B contractor

Senior AI Agent & TypeScript Platform Engineer

I build production AI agents, RAG and MCP/tool-use systems, TypeScript backends, and crypto-native execution — and operate them, not just demo them.

Previously one of the most active early contributors to elizaOS, a widely used open-source AI agent framework.

Email me GitHub Download PDF
Available for B2B contracts Croatia / EU timezone
Experience
7+ yrs
AI focus
2+ yrs
Engagement
B2B · EU

Profile

Production AI engineer for operated systems.

Senior AI agent and backend platform engineer with 7+ years of software experience and 2+ years focused on AI agents, RAG, tool-use systems, and agentic workflow automation. I build production-grade agent runtimes and TypeScript backends — and operate them: logging, tracing, retries, cost controls, queues, and admin tooling. Currently founder and lead engineer at Autonomous, building Auto (auto.fun).

Earlier, as one of the most active early contributors to elizaOS, I worked on agent orchestration, runtime behavior, MCP integrations, knowledge/RAG, and memory systems — ranked among the top of the project's open-source contributor leaderboard.

B2B engagement fit

Where I add leverage fastest.

Agent workflow delivery

Planning loops, tool routing, MCP servers/clients, RAG, memory, prompt pipelines, evals, guardrails, and human-in-the-loop controls.

Prototype to production

Logging, tracing, retries, timeouts, idempotency, cost controls, queues, admin dashboards, and production debugging.

TypeScript backend platforms

Node.js/Bun services, Convex/Next.js, realtime subscriptions, PostgreSQL, Redis, REST/GraphQL APIs, CI, and deployment workflows.

Crypto-native execution

Solana/EVM wallets, Privy auth, swaps, bridges, Hyperliquid, Polymarket, DeFi data, and execution-provider abstractions.

Experience

Recent work.

Founder & Lead AI Engineer

Autonomous — Auto (auto.fun)

2026 – present
auto.fun docs.auto.fun
  • Building Auto, an AI trading agent that executes trades across crypto, tokenized stocks, prediction markets, and DeFi from natural-language conversation.
  • Designed and operate a production agent runtime: Convex realtime backend, React frontend, Privy-managed embedded wallets, and provider abstractions for swaps, bridges, perps, and market actions.
  • Built multi-step execution workflows spanning intent parsing, wallet state, market data, protocol actions, transaction preparation, and result reporting.
  • Coverage spans 12+ chains and 10+ DeFi protocols/venues (Hyperliquid, Polymarket, Aave, Morpho, CoW, Jupiter, Enso, Uniswap, 0x), with gas sponsorship on most chains.
  • Hardened reliability: wallet identity, gas sponsorship, position reads, bridge-provider failover, stale-token recovery, and request fanout.
  • Built backoffice tooling for users, conversations, message processing, scheduled tasks, agent state, logs, and wallet data.

TypeScript, Bun, Convex, React, Next.js, Privy, Redis, Vercel, multi-chain Solana/EVM

Early Core Engineer

elizaOS (Eliza Labs)

Jan 2025 – Apr 2026
github.com/elizaOS/eliza contributor leaderboard
  • One of the most active contributors to elizaOS in its early days — a widely used open-source AI agent framework; ranked on the public contributor leaderboard as @0xbbjoker.
  • Worked across the full agent stack: orchestration, plugin architecture, runtime behavior, entity/request context, CLI behavior, logging, and production safety fixes.
  • Shipped public fixes credited in elizaOS releases, including CLI secret-leakage prevention and dev-watch rebuild-loop prevention.
  • Built and maintained major plugin surfaces: MCP, Knowledge/RAG, Memory, web search, social/channel integrations, and onchain tooling.
  • Improved plugin deployment modes, dynamic MCP tool actions, per-user MCP connections, batch embeddings, PDF parsing, long-term memory extraction, and adaptive context.

TypeScript, Bun, Node.js, elizaOS core, MCP SDK, Vite/React, RAG, memory systems, plugin architecture

AI Engineer

Orion (oriontaraban.ai)

B2B contract
oriontaraban.ai
  • Delivered an end-to-end production AI coaching agent for a client — detailed in Selected work below.

Software Engineer

Thentia

Oct 2023 – Nov 2024
  • Backend engineer on a regulated SaaS platform, focused on Node.js service development and maintainability.
  • Led backend work to refactor legacy PHP toward Node.js services, improving performance and long-term maintainability.
  • Worked with Redis, caching, microservice boundaries, and production backend workflows in a compliance-heavy environment.

Node.js, TypeScript, Redis, microservices, SaaS backend engineering

Software Engineer

Barrage

Nov 2020 – Oct 2023
  • Built backend services across fintech and crypto-adjacent systems, focused on Node.js/TypeScript microservices and real-time data.
  • Developed services for trading workflows, authentication, 2FA, KYC, crypto wallet services, notifications, and dashboard-facing APIs.
  • Worked across backend and web surfaces using Node.js, Next.js, databases, external APIs, and production support.

Node.js, TypeScript, Next.js, microservices, fintech systems, realtime data

Selected work

Representative systems.

Orion

Client delivery · Live

Production AI coaching agent that turns an expert's content library into a conversational coach. TypeScript content-to-skill pipeline, Convex state, Telegram/iMessage delivery, progressive memory, goal-lifecycle tracking, subscriptions, safety gating, and admin tooling — delivered end to end for the client.

oriontaraban.ai

Auto

My product · Live

AI trading agent that executes trades across crypto, tokenized stocks, prediction markets, and DeFi from natural-language conversation — built and operated end to end across 12+ chains (full detail in Experience).

auto.fun docs.auto.fun

elizaOS — open-source contributor

Open source

One of the most active contributors to elizaOS in its early days — a widely used open-source AI agent framework. Contributions are public on the org monorepo and the contributor leaderboard.

github.com/elizaOS/eliza leaderboard