Nikhil Singhal
Active Builds
Agentic Enterprise Search Platform
An agentic enterprise search platform built for a large enterprise client: multi-agent orchestration over RAG and custom MCP servers, reaching across a dozen-plus internal data sources, with access governed down to the row by ACLs, so people get trustworthy answers across scattered systems without ever seeing the complexity underneath.
Property Tax Appeals Platform
A property tax appeals platform that runs millions of parcels through a 15-step pipeline and 33 agent gates, turning raw assessment data into appeals accurate enough to challenge a county's own valuations.
The Intent Layer Live Beta
My AI command center, seven iterations in the making. One place that helps coordinate all aspects of your work and personal life, where you just talk and things get done. Under it: a set of AI personalities that each carry their own memory and pick up where they left off, wired to your tools and data through MCP, all running on captured intent, across both Claude and Copilot. In early beta at theintentlayer.com.
Physical AI
The hardware end of AI. Open-weight models like Qwen and Gemma running locally. The inference and orchestration platforms above them, Bedrock, Vertex AI, Foundry, NemoClaw. The harnesses and context layers that turn a model into an agent. And the physical layer where AI meets hardware: robots, vehicles, drones, surgical systems, autonomous vehicles. All of it driven by world models, reinforcement learning, and the VLA models teaching machines to act in the physical world. Deep in it. Full reference at the library.
What the Building Taught Me
Human Intelligence Partnership Charter
Four partnership patterns for human-AI systems. Fifteen positions on the map. Drop "artificial." Start building with intelligence. The charter emerged from a year of building, not from a whiteboard. Every product in the portfolio maps to one or more patterns. The overlaps create depth. The center, where all four patterns converge, is where full partnership lives.
From Instinct to Intent — Article Series
A five-part series examining the structural gap between human intent and machine execution. USPTO trademark filed (Serial 99690685). Book in progress.
Article 1: Discovering Intent: The Journey That Starts Before You Are Ready
The manifesto. The gap exists. We are all operating on instinct right now, and clarity will only come if we build the tools for the journey.
Article 2: We Are Making AI Write Code in Languages Designed for Humans. That Is the Problem.
Instance One. Programming languages were designed for human eyes. Now AI writes the code and humans spend 91% more time reviewing it. The first concrete evidence of the Intent Layer gap.
Article 3: Everyone Is Arguing About the Engine. Who Is Building the Steering Wheel?
$5 billion raised in 90 days on which AI engine to build. LeCun, Altman, Sutskever each betting on a different architecture. Nobody building the interface between human intent and machine execution. The gap is architecture-agnostic.
Article 4: Stop Calling It Artificial. Start Building WITH Intelligence.
The HIP Charter article. The word "artificial" gives you permission to skip the communication infrastructure you would never skip with a human colleague. Four patterns of partnership. The Intent Layer visible at every pattern.
Article 5: Discovering Humanity Through AI
The personal thesis. Two unplanned apologies surfaced during the book writing because the AI created a space without judgment. The insight: the value is not capability. It is safety. The absence of judgment creates the space for insight to surface.
NIST Governance Submissions
Response to NIST RFI on AI Agent Security Docket NIST-2025-0035
~5,000 word submission to the NIST Center for AI Standards and Innovation. Deployer perspective across 8 organizations, 7 industries. The core argument: governance cannot keep pace with what agents can do, and the gap is worst at the boundaries between providers.
NCCoE Response: Intent Conveyance in AI Agent Authorization
~5,900 word response to the National Cybersecurity Center of Excellence concept paper on AI agent identity and authorization. Proposes intent conveyance as the missing layer, with case studies from Spotify, Amazon/AWS, and Meta.
AI Trust Commons
Open-source governance for AI agents in the enterprise. The organizational home for the article series, the HIP Charter, and the NIST submissions.
Products
ResyncLife All four patterns
A personal command center that treats your whole life as one life. Voice capture with NL parsing, calendar sync (EventKit, ICS, OAuth), family circles, goal tracking. FastAPI + SwiftUI iOS + sync engine. Full stack shipped in a day, after a year of learning across every other product.
The center of the bloom.
Recurate Tooling + Integration
A suite of five extensions for the complete AI interaction lifecycle. Composer formats input. Annotator curates output. Copier captures conversations across 7 platforms. Connect coordinates parallel AI chats. Platform compares responses across Claude, GPT, Gemini, and Grok side by side.
ScreenTrades Integration
AI-powered trading analysis. 1,882 commits over 10 months, the origin project where the AI partnership started. 20+ real-time data APIs, confidence scores, transparent reasoning. V2.0 web + iOS shipped.
TourGraph Pipeline
Tour discovery across 136K+ experiences in 3,380 destinations. Express 5 + TypeScript backend, Next.js web, SwiftUI iOS with MapKit globe.
GitaVani Pipeline
Bhagavad Gita reader. 701 verses, 17 scholars, Sanskrit audio. iOS + Android.
ClearNews Pipeline
News aggregator. 41 sources, 13 categories, no ads. iOS + Android + Web + API.
CoPilots Launchpad Tooling
M365 test account management. Azure Trusted Signing. macOS + Windows. MIT license.
Consulting
OmSpark LLC
AI engineering and consulting practice based in Seattle. We work at the intersection of enterprise systems, AI strategy, and hands-on engineering. Our engagements range from architecture advisory to building production systems.
Book
From Instinct to Intent
A memoir about discovering human purpose through AI partnership. Not a how-to guide. Not a technology book. The story of what happens when the ground shifts at AI speed and you follow the thread until clarity arrives. HOPE + INSTINCT sections complete (~35,000 words, ~140 pages). INTENT section in progress.
Patents
Career
Three decades of engineering leadership. CTO and VP Engineering roles across Expedia ($45M savings), Imperative ($1M to $4M ARR as CTO & Head of Product), The Tour Guy (400% operational efficiency), T-Mobile, Microsoft, AT&T, Hitachi Consulting (200+ engineers led). Compliance deployed across telecom, travel, fintech, healthcare, education, green energy, and IoT. 114 LinkedIn recommendations.
Profiles
Archive
My Brave Son
Shivam's leukemia journey. A father's documentation of his son's fight and recovery.
Programming Interviews
Interview preparation resource built during the Microsoft years.
Maa Tribute
Video tribute.