Claude AI Specialist — Agentic Workflows

About Avestix Technologies Avestix Technologies is the AI and infrastructure arm of the Avestix Group — an AI-native multi-family office and investment platform. We build proprietary agentic tools that replace manual workflows across outreach, CRM management, due diligence, and client operations. We move fast, work in Python, and use Claude as our primary AI engine. If you live and breathe the Anthropic stack, this is your environment. Role Overview We are looking for a Claude AI & Agent Security Specialist on a contract (1099) basis to build, extend, and harden agentic workflows powered by Claude. You will work directly with the founder on real, production-grade tools — not demos. Current active projects include an AI-driven outreach agent, document automation pipelines, and AI-assisted deal flow and due diligence tooling. You should be comfortable reading existing code, diagnosing issues, extending agents, and writing clean Python. Claude Code proficiency is a meaningful differentiator. Critically: you understand the attack surface of agentic systems — prompt injection, runaway agents, malicious code injection, SVG-based attacks — and you know how to build defenses in. OpenClaw familiarity is a strong signal. Key Responsibilities ·      Extend the AI outreach agent: add Gmail reply detection, live web research per prospect, and HubSpot sync ·      Build and refine Claude prompts for email personalization, document drafting, and analysis tasks ·      Design and implement multi-step agentic pipelines using the Anthropic SDK and tool-use patterns ·      Integrate Claude workflows with external APIs: HubSpot, Gmail, SendGrid, Google Drive ·      Debug and optimize existing Python scripts in the Avestix codebase ·      Build lightweight Streamlit dashboards for pipeline visibility and agent monitoring ·      Audit deployed agents for security vulnerabilities: prompt injection vectors, tool misuse, runaway loops, data exfiltration risks ·      Stay current on the agentic security threat landscape—follow and apply emerging research (OpenClaw, etc.) ·      Document all work clearly — the founder uses Claude Code as primary IDE Requirements ·      Hands-on experience with the Anthropic Claude API — tool use, multi-turn conversations, structured outputs ·      Proficiency with Claude Code (the CLI) — you use it daily, not as a novelty ·      Strong Python skills — you can read, debug, and extend existing code without needing everything explained ·      Experience building agentic / automation pipelines (not just chatbots) ·      Working knowledge of agentic AI security: prompt injection, malicious code injection, runaway agents, SVG attacks, MCP/tool-use attack surfaces ·      OpenClaw awareness—you follow the agentic security space and can translate threat research into defensive implementation ·      Familiarity with at least one: Gmail API, SendGrid, HubSpot API, Google Cloud ·      Self-directed and asynchronous—this is a remote contract role with clear deliverables, not micromanagement ·      Bonus: experience with Streamlit, openpyxl, python-docx, or financial data pipelines Current Projects You Will Touch ·      AI Outreach Agent: Claude-personalized emails, Day 7 / Day 14 follow-up automation, Gmail/SendGrid sending ·      Document Automation: Auto-generate Word docs (proposals, DD memos, client letters) from structured data ·      Reply Monitor: Gmail polling to auto-detect replies and update outreach state ·      HubSpot Sync: Push CRM stage updates to HubSpot via API as relationships progress ·      Streamlit Dashboard: Live view of outreach pipeline — who replied, who is due for follow-up, top prospects Compensation & Structure ·      Type: Independent Contractor (1099) ·      Rate: Negotiable based on experience—project-based or hourly ·      Location: Fully remote ·      Hours: Flexible—deliverable-focused, not hours-focused To Apply Include links to any relevant work: GitHub, Claude-powered projects, or agentic pipelines you have built. If you have done any security research on agentic systems or follow the OpenClaw threat landscape, tell us specifically what you have found or fixed. Show us what you have built — that is more valuable than a resume. No recruiters.