HILAY
TRIVEDI.
$
I build multi-agent AI systems and the infrastructure that makes them trustworthy in production.
I build AI systems
you can trust.
I'm an AI systems engineer building governed multi-agent systems, agent memory, and AI security tooling that hold up in production. My work centres on the infrastructure that makes autonomous agents trustworthy — memory layers, layered security gates, GraphRAG, LangGraph orchestration, and MCP.
That production instinct comes from 5.5+ years shipping at enterprise scale. At rtCamp, a WordPress VIP Gold Agency, I built and led real systems for Canada's largest media house and U.S. fintech firms — proof I've delivered high-traffic, high-stakes software end to end, not just prototypes.
I'm also a WordPress Core contributor and part of the initial Core AI team, where I proposed the Unified AI Management Layer. I own projects end-to-end: architecture, design decisions, delivery, and post-launch hardening.
Founded two businesses; built two products end to end. I move comfortably from technical consulting and solution architecture through to hands-on delivery.
Where I've worked
Senior Software Engineer · AI Engineer
- ▸Built an internal AI engineering framework now adopted by 80+ engineers, standardising how teams design, govern, and ship AI features.
- ▸Part of the initial WordPress Core AI team; proposed the Unified AI Management Layer to standardise AI integration across Core.
- ▸Designed governed multi-agent systems and agent-memory tooling with layered AI security gates, applying production-grade reliability and safety standards.
- ▸Growth Engineer — owned clients end-to-end: technical consultancy, business analysis, proposals, quotations, timelines, Scope of Work, and upselling.
- ▸Developed high-performance dynamic Gutenberg blocks and led frontend performance for Canada's largest media house — one of the country's highest-traffic publishing platforms.
- ▸Technical Lead for a top-rated U.S. futures prop trading firm: system architecture, sprint planning, hands-on development, code review, and post-launch support.
- ▸Led a cross-functional team of 8 engineers to deliver 200+ landing pages in 6 weeks via a reusable PHP/ACF component system, driving a 32% sales increase and 200% daily visitor growth.
Full Stack Developer
- ▸Led the full backend migration of a .NET platform to WordPress VIP — custom PHP plugins, WooCommerce integration, and RESTful APIs — resulting in a 34% sales increase post-launch.
- ▸Built an in-house Gutenberg block library in PHP and React.js (server-side rendered, reusable across projects), significantly improving team development velocity.
- ▸Contributed to WordPress Core, Meta, Polyglots, and other WordPress.org domains.
- ▸Debugged and modernised multiple legacy PHP codebases, integrating them with modern WordPress APIs and third-party services while preserving core business logic.
- ▸Delivered technical consultation directly to clients on architecture decisions, aligning backend solutions with business timelines and budgets.
Data Science Intern
- ▸Completed a 6-month Data Science internship gaining hands-on experience in AI, Machine Learning, and Deep Learning.
- ▸Worked on real-world datasets and built predictive models using Python, applying ML techniques to solve practical problems.
Software Engineer Intern
- ▸Developed a full-scale module for a company management system using the MERN stack with a team of 4 developers.
Machine Learning Intern
- ▸Built a cutting-edge web scraper to extract LinkedIn profiles and identify professionals matching specific job descriptions using ML-driven relevance scoring.
Things I've built
All repos on GitHub →A production multi-agent system where a supervisor plans, routes, and coordinates specialised agents over a shared memory layer, with guardrails on every output. Built to make autonomous agents predictable, observable, and safe to run in production.
Agent memory layer
Persistent, governed memory for autonomous agents
Agents forget context across turns and sessions, re-fetch the same knowledge, and drift without a durable, queryable store — making them unreliable in long-running production workflows.
A layered memory service: short-term working memory, episodic session memory, and a long-term semantic store backed by vector + graph retrieval (GraphRAG). Writes pass through summarisation and relevance scoring; reads are scoped per agent and per task.
- ▸Separate working / episodic / semantic tiers so recall cost scales with need, not history size.
- ▸Graph + vector hybrid retrieval to keep both relationships and semantic similarity queryable.
- ▸Deterministic write policies (summarise, dedupe, score) to stop unbounded memory growth.
Case study — architecture and design decisions only. No client code.
AI code security gate
Layered enforcement before AI-generated code merges
AI-generated code can introduce vulnerabilities, secrets, and policy violations that slip past a single review. One check is not enough for production merges.
Defence-in-depth: each commit passes through stacked gates — static analysis, policy/rule enforcement, an LLM security review, then human approval — and no stage can be skipped before merge.
- ▸Fail-closed gates: a failure at any layer blocks the merge by default.
- ▸LLM review is one layer, never the sole authority — deterministic checks run first.
- ▸Every decision is logged and attributable for audit.
Case study — layered enforcement design. No client code.
Advoksha
AI-powered legal research terminal
Multi-agent AI legal assistant for attorneys. Features a Supervisor, Researcher, Linguistic Hub, and Drafter agent architecture. Integrates OCR & translation, contract generation, and High Court/Supreme Court research grounding.
LessonBuild
AI lesson & curriculum builder
Turns a topic or syllabus into structured, standards-aligned lessons — objectives, activities, assessments, and materials — generated and refined through an AI pipeline that keeps educators in the loop.
CreatorNexus AI
Research intelligence platform
Full-stack AI research SaaS built solo. Visualises research as an interactive causal knowledge graph with multi-source verification, defamation detection, and sponsor-safe content checks. 168+ commits in production.
Giving back to the
community
WordPress.org profile →Core AI Team
Core AIPart of the initial WordPress Core AI team; proposed the Unified AI Management Layer to standardise AI integration across Core.
Gutenberg
ContributorPRs merged into WordPress/gutenberg: block colour support updates and embedded preview block improvements.
RevisionBuster
Plugin AuthorPlugin author — published on WordPress.org with full WP-CLI support and admin UI.
Translation Contributor
GlotPress150+ translations for WooCommerce, Jetpack, and WordPress core in Gujarati.
Built By WordPress (BBWPC)
Open SourceContributed PHP theme templates, block patterns, and SSR blocks to the global BBWPC open-source initiative.
Community
WordCampSpeaker, WordCamp Asia 2026 (Core AI UX for Ability Explorer); Demo, WordCamp Ahmedabad 2025. Active WordPress Slack member.
Writing &
talks
theworldtechs.com →The World Techs
BlogMy publication on AI systems, agents, and enterprise engineering.
WordCamp Asia 2026
TalkSpeaker — Core AI UX for the Ability Explorer. Talk and slides coming soon.
Posts & notes
WritingDeep dives on agent memory, AI security gates, and GraphRAG. More posts landing soon.
Let's build something
extraordinary.
Open to Forward Deployed, Founding AI, and AI Platform roles. Remote, worldwide. Also available for WordPress VIP architecture and AI tooling consulting.