I build production AI agents, LLM apps and voice AI that survive real users. 9,400+ Upwork hours, 100% Job Success, Top Rated Plus, and most clients keep me for months.
Most AI demos work once. I build the version that runs at 2am with nobody watching: retries, timeouts, rate limits, token cost, silent failures, and the edge case nobody specified.
I write the code myself and work directly with your team. No account managers in between.
How I work (AI-native, human-owned)
I use Claude Code / Cursor every day to plan, build and review. I break features into scoped tasks for coding agents, review every line they produce, and write the parts where direct coding is faster or safer. I test against the running app, not just the type checker.
Recent work
🧠 Enterprise AI Agent Orchestration Platform
Platform for building and running AI agents across a company's tools: 20+ integrations, a visual agent builder with code editor and templates, AI tagging and summarization, real-time dashboards, and a Chrome extension. Self-hosted option for data-residency requirements.
🤝 Agentic AI for Talent Sourcing (Moonhub, since joined Salesforce; covered by Forbes, Fortune, TechCrunch, WSJ)
Multi-agent pipeline that qualifies and scores millions of candidate profiles, runs personalized outreach at scale, and hands off to humans with escalation logic.
📞 Real-Time AI Voice Agent for Outbound Booking
Calls businesses and books appointments end to end. Sub-second latency, barge-in handling, WebSocket live monitoring, and Stripe billing.
Next.js, Node.js, Twilio, ElevenLabs, AWS
☎️ 24/7 AI Receptionist for Healthcare, Legal and Real Estate
Handles calls, SMS and web chat. Two-way Google/Outlook calendar sync, reminders that cut no-shows, English/Spanish, lead scoring, and CRM sync. AES-256 encryption, SOC 2 infrastructure, TCPA compliance.
Node.js, Retell AI, Twilio, OpenAI
🎤 Conversational AI Interview Platform
Lifelike mock interviews with transcription, scoring and written feedback on pace, filler words and tone. Includes a cohort admin dashboard and passwordless access on web, mobile and landline.
🏥 Patient Scheduling & Intake Automation (healthcare, NDA)
Automated scheduling and intake on top of messy legacy healthcare data and systems.
What I own in production ✅ AI agents and tool/function calling, multi-agent workflows, human-in-the-loop approvals ✅ RAG, vector search, structured JSON outputs, evals and regression tests for LLM behavior ✅ LLM cost and latency control: model routing, caching, token limits, fallbacks ✅ Real-time voice AI: inbound/outbound calling, interruption handling, latency tuning ✅ Full-stack SaaS: Next.js, React, TypeScript, Node/NestJS, Python, Postgres/Supabase, Docker, AWS ✅ Integrations: CRMs, calendars, Twilio, Stripe, webhooks, n8n
Swaleheen J. earns an estimated $5.5k/mo. That's 4.7× the typical freelancer and more than 99.62% of everyone we track.
Good fit if you need someone to own a feature from a vague goal to tested and live, and to tell you when a simpler approach would work better.
Available 40+ hrs/week with overlap into US Eastern hours.