04 / About / Siddharth Nilesh Patel
The product is part of the business.
I’m Sid Patel. I co-founded Giggr, run its operations and build products. I’ve worked on a client’s platform architecture, made our own hiring decisions and followed delivery through to payment.
- Siddharth Nilesh Patel
- Founder-operator, AI systems builder
- Bengaluru, India
Two ways into the same problem.
I grew up around a family business in Chennai and studied computer science at Sri Venkateswara College of Engineering. I’m still drawn to both sides of a problem: how a system works, and how a business makes it worth building.
A larger share of the problem.
My early work was technical: mobile interfaces at Hatch, recommendations across more than 90,000 documents at Paperflite, and a retail price-elasticity project with Nextail during my data science programme in Madrid.
Co-founding Giggr in 2021 widened the responsibility. We used paid client work to help fund product development. As COO, I directly managed a distributed team of 25+ and took on hiring, finance, partnerships and delivery across India and the US. A product decision also became a staffing decision, a delivery commitment and a bill we had to be able to pay.
Follow the experience timelineThe number needed a reason.
Marriott’s brief for its India technology centre proposed growing headcount from 35 to 500 over five years without explaining the number. I challenged it. Through Giggr, I led workforce strategy work that included interviews with technology tower leads, provider evaluation and the case for the Hyderabad Technology Accelerator. We needed to establish what the team would do before deciding how large it should be.
The engagement also took me into platform architecture: mapping nine business processes across four levels, through review and sign-off. I created CodeFest to give Marriott associates a structured way to develop and demonstrate ideas. These were different responsibilities within the same engagement: test the staffing plan, make the processes reviewable and give people a place to build.
What would let someone else examine this?
That question now runs through my independent AI work. I created Agent Maturity Compass to put agent behaviour on the record: the actions taken, the controls applied and the evidence behind a maturity score. It gives an evaluator something to inspect beyond the agent’s explanation of itself.
Room for the questions without a brief.
Can we reduce storage and compute while checking that the result still meets its requirements? URP explores that question, with local proof so far. Can an AI answer retain evidence of its sources and their contribution? RDLLM links claims to supporting passages and checks rights before producing instructions for allocating value to sources. It is a beta and does not execute payments.
Other questions are more speculative. Kaprekar Geometry connects exact mathematics to a still-unvalidated hypothesis about model reliability. My AI-assisted Indus script study tests statistical patterns and published readings against null models; it is an unreviewed preprint and claims no decipherment. Closer to everyday life, ADHD Life OS is a pre-release tool for executive-function support, and Style Second Opinion is a private prototype for volunteer style conversations between verified women.
Read the lab notebookThe organisation around the technology.
I co-write Giggr Glance Over with Giggr’s founding team, for a readership of 10K+ subscribers. The essays examine capability, enterprise design and AI: what changes when people can do more, and what still gets in their way. They sit alongside the products because a new tool raises questions about how an organisation should work.
Browse the writingHands on, with responsibility for the result.
I write TypeScript and Python. The work ranges from AMC’s agent runtime and evaluation machinery to applied ML and statistical analysis. URP also uses Rust and Go; parts of the Kaprekar mathematics use Lean 4 and Mathlib. I choose the language around what needs to be implemented or checked.
I also direct AI coding agents, including on KhelPath and Style Second Opinion. I define the task, inspect the implementation and decide what evidence is enough to call it working. The checks and the claims about the result remain my responsibility.
A record you can follow.
The case studies explain my role, the decisions and the result. The lab records tests and open questions. Status labels tell you whether you are looking at a release, a demo, a prototype or research in progress. The colophon explains how this site was made.
Read the colophon