Aayush Goel

I started shipping software commercially before I started university. Between 2020 and 2022 I founded and ran SoloNodes.net, a game and web server hosting business, while still in high school — writing the client dashboard in Node.js and EJS with points-based billing and automated provisioning, growing a Discord community past 10,000 members, coordinating freelancers, and eventually selling the business.

I studied BCA (Hons.) in Data Security at the Symbiosis Institute of Computer Studies and Research (SICSR), Symbiosis International University in Pune, from August 2022 to May 2025, graduating with a CGPA of 8.61. I won the institute's internal hackathon in December 2024, leading a team to a working solution in 24 hours.

I joined Mulltiply in January 2025 as a Full Stack Engineer and was promoted to Engineering Lead in February 2026. Mulltiply is a WhatsApp-first B2B commerce platform: businesses sell to their buyers inside WhatsApp and Instagram rather than through a conventional storefront, which turns messaging APIs into the primary transaction surface and makes latency, idempotency, and multi-tenancy the hard problems.

What I actually build

• Multi-tenant WhatsApp commerce infrastructure. A routing layer that lets a single WhatsApp business number serve a group of stores, across 8+ Node.js microservices on Kubernetes with ArgoCD. The rules live in a 4×4 visibility/registration matrix, illegal combinations are rejected server-side, and every failure path degrades to the behaviour that existed before the feature shipped.
• Latency work users feel. Chatbot replies went from ~10s to ~1–2s: the subscription row moved into a cache, replacing 12+ serial reads with one lookup, six more collapsed into a single Promise.all, and a ~500ms language-detection call moved off the critical path. Separately, catalogue search went 281ms → 29ms once I found a partial index whose predicates a refactored join no longer satisfied, so Postgres was ignoring it entirely.
• Bidirectional Tally ERP synchronisation. Two-way master and transaction sync with automated invoice delivery over WhatsApp, and idempotency guards on every operation so retries never double-post — eliminating manual cross-system accounting.
• Instagram chatbot onboarding, self-serve. Merchants fill a structured form that generates a custom AI system prompt, preview the bot live in-browser, then connect their Instagram account and go live automatically through the Instagram Graph API with zero manual intervention from the team.
• Multi-agent LLM automation. A five-agent marketing system, shipped solo — specialised CMO, Content, Graphic Designer, FB Handler, and Analyst agents take a campaign brief to a live Meta ad, with DALL·E 3 creative generation, human approval gates, and full LLM call audit logging for tokens, latency, and cost.
• Conversational AI decisioning. I designed the chatbot decisioning architecture — intent classification, cart extraction, price negotiation, multilingual handling — and directed its implementation in LangGraph and Python on the OpenAI API. The earlier pipeline I built myself handles 18+ intents across English, Hindi, and Hinglish.

How I work

I'm a lead who still writes code. Day to day that means architecture decisions and code reviews for a three-person frontend/backend/AI team, running a full intern hiring cycle including group discussions and technical interviews, and then going back to the branch. I prefer boring, observable systems — Prometheus and Grafana on everything, queues over cron, idempotency keys over hope — and I treat an integration as unfinished until it survives a partner API behaving badly.

What I'm into

Conversational commerce and messaging-API architecture; ERP and accounting-system integration, particularly Tally and TDL; multi-tenant SaaS data modelling; retrieval-augmented generation and multi-agent orchestration with LangChain and LangGraph; and workflow automation with n8n.