# Hire Ayush Goyal — Freelance Forward Deployed Engineer > I embed with your team, find where AI actually pays off, and ship it into your production stack — agents, voice, RAG and integrations. Deployed, not demoed. **Status:** Taking freelance forward-deployed engagements **Location:** India — remote worldwide **Book a free 30-min scope call:** https://cal.com/ayuugoyal **Email:** ayushgoyal8178@gmail.com **Website:** https://www.ayuugoyal.tech **GitHub:** https://github.com/ayuugoyal **LinkedIn:** https://www.linkedin.com/in/ayuugoyal/ **X:** https://x.com/ayuugoyal **Lab:** https://www.rndynamolabs.com/ (RnDynamos Labs — Physical AI) ## What is a forward deployed engineer? A forward deployed engineer (FDE) is a software engineer who embeds directly with a customer's team to make a product or technology work in their real environment. Instead of handing over a spec, an FDE learns the workflow from end users, writes the integrations, and ships to production. For AI, that means taking a model from demo to something your team uses every day. ## How an engagement runs 01. **Scope call** — 30 minutes at cal.com/ayuugoyal. You describe the problem; I tell you straight whether AI is the right answer and what it would take. 02. **Embed & discover** — I get access to your stack and time with the people doing the work. The real requirements live with users, not in the brief. 03. **Ship to prod** — A working pilot early, in your environment, then hardened for production: auth, monitoring, evals, fallbacks. 04. **Hand over** — Documented and handed to your team — or kept on a retainer if you would rather I keep running it. ## Engagement shapes - **Deployment sprint** — One well-defined use case, taken from zero to production over a fixed few weeks. - **Pilot to production** — You already have a demo that works on someone's laptop. I make it survive real users. - **Embedded retainer** — Part-time FDE on your team, shipping and maintaining AI features month to month. ## What gets deployed ### Forward Deployed AI Engineering I join your team for a scoped engagement, learn the workflow from the people who run it, and take one AI use case from messy reality to production — inside your stack, under your security rules. - Discovery with the actual end users, not just the sponsor - Working pilot in your environment within the first weeks - Production rollout: auth, monitoring, fallbacks - Handover docs and a team that can run it without me Proof of work: Machine Maintenance Bot (live on a factory floor), SiteOS (live for an architecture studio) ### AI Agents, Chatbots & WhatsApp Bots Assistants that answer from your own content instead of hallucinating — on your site, in Slack, or on WhatsApp where your customers already are. Multi-step tool use and a clean handoff to a human. - Embeddable support & sales chatbots - WhatsApp Business API agents - Multi-step agents with tool calling - Guardrails, evals and fallback handling Proof of work: Machine Maintenance Bot, SiteOS, Chatter AI, LawGPT ### AI Voice Agents Phone agents that hold a real conversation in Hindi and English — qualifying leads, booking appointments and handling inbound calls without a queue. - Inbound & outbound calling agents - Telephony integration (Twilio / Plivo) - Call transcripts, summaries and CRM sync - Latency and interruption tuning Proof of work: Mona @ RnDynamos Labs, bolna-ai/bolna, ArduPilot Assistant ### Enterprise Integrations Most of forward-deployed work is plumbing. I connect AI to your CRMs, drives, databases and SaaS tools with auth, retries and error handling that survive production. - SaaS & internal API integrations (OAuth 2.0) - Notion, SharePoint, OneDrive & Drive connectors - Slack, WhatsApp, Gmail & Google Workspace - Webhooks, ETL jobs and rate-limit-safe sync Proof of work: SharePoint ticketing in Machine Maintenance Bot, SiteOS SharePoint archive, archestra-ai connectors ### RAG & Knowledge Systems Answers with receipts. Retrieval over your documents, wikis and drives, wired into whichever model you want, with citations your users can check. - Embeddings and vector store design - Chunking, retrieval tuning and reranking - Citation output and grounding checks - Provider-agnostic model layer (Claude / OpenAI / Gemini) Proof of work: LawGPT, archestra-ai connectors, Chatter AI ### MCP Servers & Agentic Workflows Custom Model Context Protocol servers and multi-agent flows that give AI real, permissioned access to your systems — with human approval gates where it matters. - Custom MCP servers for your APIs and data - Planner / executor and supervisor agents (LangGraph) - Tool permissions, audit trails, human-in-the-loop - Claude Code / Cursor workflow setup for your team Proof of work: AI DevOps Agent, slack-claude, QuickDocs, archestra-ai ### LLMOps, Evals & Guardrails The difference between a pilot and production. Tracing, eval suites, prompt versioning, and the cost and latency work that keeps the AI feature affordable. - Tracing and observability (LangSmith) - Eval suites and regression tests for prompts - Cost, latency and token optimisation - PII handling and prompt-injection defence Proof of work: Production AI at Loadshare Networks, archestra-ai, Chatter AI ### Workflow & GTM Automation The boring, expensive manual work — gone. n8n pipelines, internal bots, lead enrichment and outbound that run quietly in the background. - n8n and custom pipelines across your stack - Lead scraping & AI enrichment - Personalised outbound with caps & suppression - Reporting and spreadsheet automation Proof of work: Lead Enrichment Tool, Cold Outreach Engine, Cloudflare Tunnel + n8n @ Data Alt Dynamics ### Physical AI & Robotics Integration AI that touches hardware. ROS2 control stacks, robot arms, IoT sensor networks and live factory dashboards — from firmware all the way up to the browser. - ROS2 control stacks and embedded firmware - Robot arm integration (6-DOF, SCARA) - IoT sensor networks and OEE dashboards - Voice and LLM control of real machines Proof of work: BCN3D Moveo, SCARA Robot, Sensor Dash, UK Design 6450987 ## Where Ayush has deployed - **Machine Maintenance Bot** — Production WhatsApp bot running a factory's machine-breakdown workflow. Operators report breakdowns through WhatsApp Flows with photos, supervisors accept and dispatch technicians, and every ticket is tracked in SharePoint — with optional 5-Why root-cause analysis and a nightly PDF report with Gemini-written analysis. Redis locks and idempotent intake keep parallel supervisors from double-acting. Deployed as a freelance forward deployed engineer; live in production. _Stack: Node.js, Express, WhatsApp Cloud API, WhatsApp Flows, SharePoint / Graph API, Redis, Gemini, Puppeteer, Docker._ - **SiteOS** — Project, drawing and site-update platform for an architecture studio. Site engineers send photos, video or notes on WhatsApp; SiteOS works out the project and floor, groups a 10-minute window into one update, stitches media into a single video, and pushes a ticket to the project manager — plus a daily report page and SharePoint archive. The office gets a dashboard with DXF plans as clickable layouts, versioned drawings and strict role-based access. Deployed as a freelance forward deployed engineer; live in production. _Stack: Next.js, React, PostgreSQL, Drizzle, BullMQ, Redis, ffmpeg, WhatsApp Cloud API, SharePoint, Docker._ ## FAQ ### What is a forward deployed engineer? A forward deployed engineer (FDE) is a software engineer who embeds directly with a customer's team to make a product or technology work in their real environment. Instead of handing over a spec, an FDE learns the workflow from end users, writes the integrations, and ships to production. For AI, that means taking a model from demo to something your team uses every day. ### Can I hire Ayush Goyal as a freelance forward deployed engineer? Yes. Ayush takes a small number of freelance forward-deployed engagements, fully remote and worldwide. Book a free 30-minute scope call at cal.com/ayuugoyal or email ayushgoyal8178@gmail.com with what you are trying to ship. ### How is a forward deployed engineer different from an AI consultant? A consultant typically leaves you with recommendations. A forward deployed engineer leaves you with running software: code in your repositories, integrations with your systems, and monitoring in production. The advice is a side effect of doing the work. ### How do you freelance while working full-time as an AI Engineer? Ayush works full-time as an AI (Harness) Engineer at TAP Innovations and takes a limited number of scoped freelance engagements on top. Timelines and availability are agreed up front on the scope call, so there are no surprises mid-build. ### Can you work inside our existing stack and security rules? That is the point of forward-deployed work. Ayush builds provider-agnostic, with Claude, OpenAI or Gemini selected by config, and has shipped across Next.js, FastAPI, Flask, Express, Postgres, MongoDB, Docker, AWS and Azure. Ayush works within your access controls rather than around them. ### What does a typical engagement look like? It starts with a scope call, then an embed-and-discover phase with your users, a working pilot in your environment early, a production rollout with monitoring and evals, and a clean handover. Shapes range from a fixed deployment sprint to an ongoing embedded retainer.