SP
Open to conversations [ Let’s talk ]
00 / MMXXVI SID PATEL / SELECTED WORK

Siddharth Nilesh Patel

Business judgment. Built into systems.

I’m Sid Patel, co-founder and COO of Giggr. I led our platform architecture and workforce strategy work for Marriott. I also built Agent Maturity Compass independently to examine what AI agents actually do.

Explore selected work
Sid Patel in outline, looking left with his arms crossed.
Bengaluru, IndiaBuilder / Operator
Who
Siddharth Nilesh Patel
What
Founder-operator, AI systems builder
Where
Bengaluru IST
Updated
November 2026

01 / Selected work

Decisions made.
Work to show.

View the work archive

GIGGR

01 / Co-founder & COO

Operating

2021 / Now

At Giggr, paid client work helped fund our own products. As co-founder and COO, I lead the operating work behind both: hiring, finance, partnerships and delivery across India and the US.

Read the case study

Giggr’s seven capability domains, connected through its four-stage approach to developing an idea.

Orbital stations
Clockwise from the top: Environment, Education, Health, Wealth, Mobility, Technology and Governance.
Bottom · Four-stage strip
Conceptualization, Creation, Contextualization and Commercialization.
Marriott billings
$450K+
People led
25+

MARRIOTT

02 / Workforce strategy & architecture

Delivered

Feb 2023 / Aug 2024

Through Giggr, I challenged the staffing assumptions behind Marriott’s India expansion, led workforce strategy and mapped nine business processes to its platform architecture. I also created CodeFest, giving associates a structured route from an idea to a working prototype.

Read the case study

A schematic of the technology accelerator. It represents the workforce-planning scope, not an architectural floor plan.

Left · Floor plate study
Desk banks illustrate the Hyderabad workforce-planning scope.
Right · Process study
BPMN swimlanes show the process side of the accelerator.
Planned accelerator seats
3,000
Processes × levels
9 × 4

AMC

03 / Founder & lead author

Released v1.1.1

2025 / Now

An agent’s account of its behaviour is a claim to check. I created Agent Maturity Compass to score maturity from evidence of what agents actually do, with the record behind each assessment available to inspect. v1.1.1 is released; the native agent runtime is in development for v1.2.

Read the case study

An illustrative maturity dial and evidence record. Try the dial to see the example level change.

Left · Maturity dial
Levels run from L0 to L5. L3 is the starting example.
Right · Evidence ledger
The record pairs the example level with an illustrative signed receipt. The dial control reports the selected level.
Maturity scale
L0–L5
npm release
v1.1.1

KHELPATH

04 / Product & build

Live demo

2026 / Demo

I directed AI coding agents to build KhelPath: a demo that checks eligibility, shows the evidence required and then ranks possible roles. Fictional records; partial translations.

Read the case study

An illustrative sports-pathway network. Eligibility and evidence determine which next steps are available.

Network · Illustrative routes
Connected nodes sketch possible pathways.
Passport · Sports and roles
The panel lists 201 sports and 608 roles.
Live demo · Journeys
The demo presents 18 journeys.
Sports
201
Roles
608

DMC

05 / Product & diagnostic design

Live

Giggr / Product

I built Digital Maturity Compass to help teams decide what needs attention before their next technology investment. Its 42 questions examine domain, leadership, culture, resilience and readiness separately, so an overall score does not hide where the gaps are.

Read the case study

A sample assessment across five dimensions. The uneven shape shows why a single average can hide a weak area.

Left · Assessment radar
Illustrative profile, not a measured result. Clockwise from the top: Domain, Leadership, Culture, Resilience and Readiness.
Right · Sample question
The assessment panel shows question 17 of 42.
Questions
42
Assessment
~8 min

RDLLM

06 / Author & maintainer

v1.0.0 · Beta reference

2026

I built Royalty Driven LLM to connect individual AI claims to the passages that support them. Evidence and rights checks determine whether it can produce instructions for allocating value to sources. The beta records those instructions; it does not execute payments.

Read the case study

A claim links to its source passage and a signed attribution receipt. Rights checks govern allocation instructions, which do not execute payments.

Left · Evidence span
The highlighted line marks the source passage.
Center · Claim and hash
The claim is linked to the source evidence and its hash.
Right · Signed receipt
The receipt records attribution.
Internal conformance cases
187
White paper pages
17

URP

07 / Author & lead builder

v0.1.0 · Local proof

2026

I built Universal Reduction Plane to check whether a smaller file or cheaper computation still meets explicit requirements. If it fails, the system returns the original. The release demonstrates that workflow locally; production savings remain unmeasured.

Read the case study

A reduced candidate must pass its preservation check. If it fails, the system returns the original baseline.

Left · Baseline
The original result remains available as the fallback.
Center · Preservation contract
Check the reduced candidate against the contract.
Right path · Verified output
A passing candidate goes forward. The dotted loop returns to the baseline on failure. Both outcomes are recorded in the manifest and ledger.
Open-source release
v0.1.0
Technical report pages
16

Kaprekar Geometry

08 / Independent researcher

Manuscript in preparation

2026

In the four-digit decimal routine, 9,990 starting states reach 6174 within seven steps. I found a smaller way to describe those paths without losing the exact counts. Some counting lemmas are formalized in Lean. The model-uncertainty idea is a separate hypothesis, with a benchmark designed to test it.

Read the case study

The four-digit sort-and-subtract routine reaches 6174. This exact arithmetic does not validate the separate AI-reliability hypothesis.

Rule
Subtract the ascending digits from the descending digits and retain four digits.
Connected states
3524 → 3087 → 8352 → 6174.
Highlighted fixed point
7641 − 1467 = 6174. Four identical digits instead give 0000.
Configurations checked
49
Manuscript pages
17

ADHD Life OS

09 / Founder & product designer

Early build · Pre-release

2026 / Now

When someone loses momentum, the tool should help them decide where to pick up. I designed ADHD Life OS around choosing a next action that fits the time and energy available. My five-second target gives the design a test it still needs to pass with users.

Read the case study

A concept of the next-action flow: choose by time and energy, then enter a focus sprint. This is a schematic, not a product screen.

Left · Open loops
The stacked rows represent pending tasks.
Center · Next action
Choose one action that fits the time and energy available.
Right · Focus sprint
Start one bounded work session for the chosen action.
Next-action design target
5 sec
Core workflows offline
Local-first

Style Second Opinion

10 / Founder · Ideation & prototype

Private prototype · Not launched

Oct 2026 / Now

Could someone deciding what to wear get a useful second opinion in a five-minute call? I defined a free service connecting verified women with volunteers, then directed a coding agent to prototype the call, matching and moderation flows. Both people can opt to extend the call. The prototype is private.

Read the case study

The proposed five-minute conversation can extend to ten only when both participants agree. This depicts the private prototype’s design.

Left · Requester
One person asks for a perspective.
Center · Five-minute timer
Both people can agree to extend the conversation to ten minutes.
Right · Volunteer reviewer
The match can be random or based on taste.
Proposed call
5 min
With mutual extension
10 min

Indus Script

11 / Independent researcher

Unreviewed preprint

Oct 2026 / Now

I led an AI-assisted study to see which Indus signs recur in particular positions and which patterns hold across catalogues. I tested published readings separately against chance. The result is a structural analysis with public methods and an unreviewed preprint. I make no decipherment claim.

Read the case study

A schematic of corpus comparison and positional analysis. The marks and bars are illustrative, not transcriptions or measured results.

Left · Corpus grid
The marks are abstract placeholders.
Right · Positional distribution
The illustrative bars show how positions could be compared against null models.
Preprint pages
54
Positional signs replicated
28

02 / Capabilities

The business.
The people. The system.

Business & organisation

AI systems

Product & operations

03 / Evidence

The scale of the work.

DATA / 01
$450K+

Giggr billings from the Marriott engagement

Profile · billing period

The supplied professional profile reports $450K+ and approximately 3,100 billed hours, February 2023 to August 2024. It describes my responsibility from approvals through invoicing and collections; the underlying invoices are private.

Read the engagement
DATA / 02
25+

People I directly managed across Giggr’s India and US team

Profile · team scope

The Giggr COO entry in the supplied professional profile records a distributed team of 25+ under my direct management. The figure describes my management scope, not total company headcount.

Read my role at Giggr
DATA / 03
3,000

Planned capacity of Marriott’s Hyderabad Technology Accelerator; I led Giggr’s workforce strategy work

Profile · planned facility capacity

The supplied professional profile records the facility’s planned 3,000-seat capacity and my workforce strategy role through Giggr. Marriott announced its opening in October 2025, after our billed engagement ended. This is capacity, not a hiring count.

Read the workforce strategy case
DATA / 04
110+

Associates in the first CodeFest I created and led: 32 teams

Profile · first edition

The supplied professional profile records 110+ associates in 32 teams at CodeFest 01 on 11 May 2023. I created and led CodeFest through Giggr; the program ran across three editions.

Read the CodeFest context
DATA / 05
v1.1.1

Released version of Agent Maturity Compass on npm

Public npm release

The published agent-maturity-compass package provides CLI, Studio, API and MCP interfaces. The native agent runtime is v1.2 work in development, separate from this release.

View the npm release
DATA / 06
10K+

Readers subscribed to the newsletter I co-write with Giggr’s founding team

Profile · newsletter

The supplied professional profile records 10K+ subscribers to Giggr Glance Over on Substack. I co-write it with Giggr’s founding team; the figure describes the shared publication’s readership.

Visit the shared publication

04 / The lab

Some questions
need an experiment.

Can an AI answer carry its sources with it? Can we use less compute and prove what we preserved? Can a mathematical pattern tell us anything useful about uncertainty? The lab is where I give those questions a test, and publish how far the answer gets.

Open the lab

05 / The line

The work changed.
The scope grew.

Full timeline

06 / Writing

What changes
when the tools change?

Giggr Glance Over / 10K+ subscribers

I co-write Giggr Glance Over with our founding team. Our subject is the organisation around the technology: how it learns, makes decisions and creates value when its tools change faster than its habits.

Subscribe on Substack
All essays, papers and talks

07 / Recognition & engagements

The wider circle.

Running Giggr has meant working with enterprise teams, university leaders, public institutions and sports organisations. This record includes delivered work, signed agreements, teaching, proposals and conversations.

2024 / Received by Giggr

Future of Work (Emerging)

2nd FICCI National HR Innovation Awards

Giggr was recognised for its work on an inclusive future of work, with an emphasis on entrepreneurship, intrapreneurship and sustainability.

FICCI’s announcement

2022 / Received by Giggr · Co-received by Sid

Best Emerging
Technology Start Up

ET Achievers Karnataka 2022

Awarded to Giggr as an emerging technology startup. I co-received it on the company’s behalf.

Giggr’s recognition record

08 / How I work

The decisions
behind the delivery.

Who should own the work? What belongs in the product? What evidence is enough to proceed? These are decisions I work through from the first brief to delivery and collection.

01 / What happens between the request and the result?

Start where the work gets stuck.

A brief tells me the intended outcome. The handoffs reveal what it will take: who has the information, who can make a decision and who is waiting. I trace that work before deciding what should change.

  • Follow one real journey through the people and systems involved.
  • Locate the waits, repeated decisions and missing information.
  • Separate a technology constraint from an ownership problem.
What comes out of it
A process map showing handoffs, decision rights and the system changes they require.

At Marriott, through Giggr, this meant mapping nine business processes across four levels and connecting them to platform architecture through review and sign-off.

Follow the Marriott work

02 / What is this plan asking us to believe?

Define the work before setting headcount.

A headcount target commits the business to salaries, management and a particular mix of skills. I work backwards from the outcome: what work is required, who can do it and what needs to happen first.

  • Identify the capabilities the business needs before accepting a staffing target.
  • Compare internal delivery and external providers against those capabilities.
  • Put the trade-off in front of the person who can decide.
What comes out of it
An operating recommendation with options, assumptions and a named decision owner.

Reviewing Marriott’s GCC brief, I challenged a five-year headcount ramp without a stated basis. The wider work included technology-lead interviews, provider evaluation and the workforce strategy behind the Hyderabad accelerator.

Read the workforce strategy case

03 / Can someone complete the task?

Build the difficult decisions into the product.

A working version exposes decisions a brief can leave open. What does a person see first? What can they do next? What happens when they do not qualify or the information is incomplete? I work through those questions in the product, writing code and directing AI coding agents.

  • Build the path from a person’s starting point to a useful result.
  • Make eligibility, permissions and missing information explicit.
  • Check the generated implementation against the intended behaviour.
What comes out of it
A working task flow with eligibility rules, permissions and explicit stopping points.

In KhelPath, eligibility is checked before roles are ranked. The demo explains the evidence a pathway requires; its records are fictional and translations remain partial.

Examine the KhelPath decisions

04 / What would make this result unconvincing?

Keep the claim within the evidence.

A successful demonstration leaves questions open. I look for a baseline, a repeatable check and the conditions under which the result stops holding. That is especially important when the system can produce a persuasive account of itself.

  • Record the behaviour being evaluated and the comparison used.
  • Exercise the controls when a request should be denied or fail.
  • State separately what the test observed and what it did not establish.
What comes out of it
An evaluation record: observed behaviour, comparison, failed checks and limits.

In AMC, a written approval policy shows what a team specified. An execution record shows whether the agent waited for approval before it acted.

Inspect the AMC approach

05 / Who owns the next step, and can we sustain it?

Connect the delivery plan to the cash.

At Giggr, a client commitment affects staffing; staffing affects cost; a delayed approval can hold up both delivery and payment. I use reviews to find those dependencies while there is still time to change the plan.

  • Use stand-ups, demos and reviews to expose blocked work.
  • Connect delivery approvals to invoicing and collections.
  • Give open decisions a responsible person and a next action.
What comes out of it
A delivery plan with owners, review points and the approvals needed to invoice.

I directly managed 25+ people across India and the US. For the Marriott engagement, my responsibility also covered the work from delivery approval through invoicing and collection.

See the operating side of Giggr

09 / Now

On my desk.

Read the Now page
Building
AMC v1.2 In development
Writing
The next Glance Over
Local time / Bengaluru
IST
Languages
English, Hindi, Gujarati, Tamil

10 / Contact

What are you
working through?

Ways to work together
Signal / Contact Direct line
LinkedIn

Site index

HomeColophonBengaluru ISTLinkedIn ↗GitHub ↗