Shoaib Akhtar

• ◆Sole engineer on the core AI product, owning end-to-end delivery of a clinical AI platform deployed in live hospital environments and working directly with doctors and admins to translate ambiguous workflows into shipped product.
• ◆Designed a microservices backend with FastAPI, Node.js, and a Go WebSocket relay supporting real-time AI workflows, with p95 latency under 250ms on core endpoints.
• ◆Built a multi-agent system on LangGraph orchestrating 5+ specialized agents for clinical task automation, with structured handoffs and tool-use for downstream EMR actions.
• ◆Shipped a production RAG pipeline on self-hosted Qdrant over hospital-specific clinical documents, with chunking and retrieval tuning that meaningfully improved answer relevance for end users.
• ◆Implemented LLM-as-a-Judge evaluation pipelines for response observability, enabling automated regression checks on prompt and model changes before each release.

• ◆Built a social-commerce mobile app (React Native + Node.js/Express) from scratch as the founding engineer, shipping core feeds, profiles, and media-heavy interactions for early users.
• ◆Deployed on AWS (EC2, S3, Lambda) for scalable media storage and event-driven processing; integrated Stripe for payments and creator monetization.
• ◆Designed REST APIs and integrated zero-shot object detection (Grounding DINO) to identify products from images, powering content-aware product recommendations.

• ◆Built a Chrome extension aggregating ticket prices across 6 ticketing platforms, surfacing the cheapest option directly inside the user's existing browsing flow.
• ◆Optimized backend APIs and PostgreSQL queries on AWS (EC2 + RDS), reducing core endpoint latency by 100ms.
• ◆Implemented a cron-based price tracker and a React-based injected UI that rendered directly inside supported ticketing sites.

• ◆Built core backend services on Node.js/Express with Redis caching, taking average API latency on hot paths under 100ms.
• ◆Shipped a no-code chatbot builder where users upload documents (Azure Blob) and get an ML-backed conversational agent, with a React/Next.js frontend for configuration.
• ◆Containerized services with Docker on Azure Container Apps and set up CI/CD with GitHub Actions, cutting deployment turnaround time meaningfully.

Autonomous AI scheduler that calls participants, negotiates availability, and resolves conflicts.

• ◆Handles concurrent calls to multiple participants, using LLM-based transcript analysis to parse vague spoken responses into structured candidate time slots.
• ◆3-phase conflict resolution engine that searches across participant slot combinations to find a feasible meeting time within a configurable scheduling window.
• ◆Event-driven pipeline with BullMQ workers and Socket.io for real-time status propagation to the frontend.

Natural-language analytics layer over GA4 + SEO data using a LangGraph multi-agent setup.

• ◆Multi-agent pipeline (GA4 agent, SEO agent, fusion agent) that classifies user intent and routes natural-language questions to the right data source.
• ◆Integrated GA4 Data API and Google Sheets crawl exports, translating free-form questions into structured API calls with automatic substitution of deprecated metrics.
• ◆Dual-LLM setup with a LiteLLM proxy and Gemini fallback, with retry logic and graceful degradation under provider rate limits.

Gmail-integrated classifier that organizes inbox into spam, promotions, jobs, social, and important.

• ◆OAuth-based Gmail access for secure retrieval of user emails without storing credentials.
• ◆MIME-aware parsing layer to extract clean email content across plain-text, HTML, and multipart messages.
• ◆Trained and deployed a classification model that buckets incoming mail into 5 categories for inbox triage.