Riya Tilwani

Building product infrastructure for live streamers and content creators, with tools that help creators manage high-pressure live moments, content workflows, and audience-facing operations with less manual overhead.

Creator products have to feel fast and lightweight while the product direction is still being shaped. The work needs strong product judgment, flexible architecture, and careful decisions about what to automate without making the creator lose control.

Owning early product engineering across architecture, backend, frontend, integrations, and workflow decisions so the startup can move quickly from founder insight to usable product.

Leading product engineering and delivery at Gryd.io across core products, client workflow systems, and technical decisions that need both strategic direction and hands-on execution.

The work spans product architecture, backend systems, deployment pipelines, Atlassian-based client workflows, AI experimentation, and coordination across a small part-time engineering team.

Keeping delivery practical and maintainable while shaping product decisions, managing developers, owning DevOps tradeoffs, and building AI-assisted workflow tools where they create real leverage.

Built a self-hostable AI assistant for enterprise teams that needed data control, internal search, speech input, and code execution without depending on a single hosted vendor.

The product had to feel like a modern assistant while staying deployable in controlled environments. The hard parts were model portability, data retention boundaries, real-time search, and safely turning assistant output into executable workflows.

Delivered a Chainlit-based assistant with open-source LLM support, PostgreSQL-backed persistence, Tavily web search, Whisper speech-to-text, and LLM sandbox execution.

Built a web app that turns user inputs and uploaded CSV files into customizable business reports with generated copy, charts, images, live previews, and PDF export.

The system needed to combine creative generation with deterministic document production. Users needed template control and version history without losing the speed benefits of AI-generated drafts.

Delivered a full-stack report builder using Grok AI for content, OpenAI for chart generation, MongoDB for versions, and Fileforge for PDF generation.

Built a subscription web app for a law firm to upload audio files and produce documents through a fine-tuned, context-aware speech-to-text flow.

The product needed legal-context accuracy, a smooth upload-to-document workflow, subscription tiers, and cloud delivery without making operations heavy for the client.

Delivered the SaaS workflow with Stripe subscriptions, React UI, Django backend, a fine-tuned Whisper flow for legal context, and AWS CloudFront distribution.

Contributed to a race data management platform for Le Mans racing operations, where teams need clean data flows, reliable views, and fast-moving operational tooling.

The hard part was not only speed. Valuable race information lived in loose Excel sheets with extremely unstructured, nuanced data, so it needed careful modelling before it could support retrieval or analysis.

Built and improved platform capabilities as an individual contributor, turning messy operational data into more usable flows for search, analysis, and production-ready delivery.