Mrunmai Patil — trustworthy software at AI speed

Mrunmai — a photograph blending into an illustrated portrait of herself, framed by a watercolour halo of green, orange, and violet.

Introduction

सर्वे भवन्तु सुखिनः

Mrunmai Patil (M) · Pune, India · works worldwide

I build trustworthy software at AI speed.

Agentic engineering and local-first AI — production-grade, secure, and shipped in days. Not vibe-coded and hoped for; reviewed, tested, and yours to run.

Ex-NVIDIA cloud & product. Founder of Leap of Faith Labs. A computer engineer and systems designer who ships fast without leaving the safety behind.

Tell me what you need built →
See offers & pricing

AI Build Sprint

From $2,000 · fixed scope · days not months

A production-ready prototype, internal tool, or local-first app — spec to deploy, tested and documented. Python/JS, React/Node, on-device AI where it fits. You own the code.

Start a sprint →

Vibe-Code Rescue & Audit

Audit from $1,500 · 3–5 days · hardening $3k–$15k

For AI-built MVPs that are 90% there — or quietly leaking. A security & architecture review (OWASP Top 10, secrets, auth, row-level security), a plain-English report with costed fixes, and optional hardening to production.

Book an audit →

Ship-It-Safely Retainer

From $2,500/mo · fractional AI engineer

For teams shipping AI code weekly with no review net. Ongoing review, security, and architecture guidance on a cadence — velocity high, risk low. CTO-level engagements to $15k/mo.

Talk retainer →

Speed of vibe-coding, discipline of engineering.

Reviewed for security

Every build checked for the OWASP Top 10, hardcoded secrets, real auth, and row-level security — the step that separates production software from a demo.

Tested & documented

Tests and handoff docs, so it survives contact with production and the next engineer who touches it.

Local-first when it matters

On-device / zero-egress options for privacy- and DPDP-sensitive work — AI features that never send data to a server.

Protocol over platform

People at the centre — the ethic I carry from Leap of Faith Labs into every build.

Zero downtime and zero critical incidents patching mission-critical apps since Jan 2020. Shipped NVIDIA's first Azure Front Door + WAF; CIS-hardened images with Packer & Terraform; DataDog observability; automated SOX monitoring & cloud cost governance.

An EdgeAI facial-recognition system (featured at GTC 2022) and NVBot, an enterprise AI chatbot — with KPIs on availability, reliability, and adoption, and a data-protection review for trustworthy AI.

Leap of Faith Labs — local-first, privacy-first AI and digital public infrastructure for education. Agentic engineering in practice: Claude skills, MCP servers, eval harnesses, on-device LLM apps.

AWS-certified · BE Computer Engineering, Savitribai Phule Pune University (8.88/10) · Georgia Tech — HCI · Stanford GSB — Design Thinking

  • Claude Code
  • Cursor
  • MCP / agents
  • on-device LLMs
  • Python
  • JavaScript
  • React
  • Node
  • AWS
  • Azure
  • Terraform
  • DataDog
  • local-first / DPDP

Leap of Faith Labs — the human layer on India's digital public infrastructure. Learning, culture, and care for the commons: made visible, valued, regenerative.

One spine, end to end: activity → evidence → verification → impact unit → regeneration.

Approach

The method and the spine — how a claim earns its keep.

Programs

After the Bell (pilot live) · Aikā ऐका (in build) · AI for ALL (cohorts opening).

Showcase

The lab — what we're building, in the open.

Commons

Open artifacts under CC-BY-SA 4.0. Use, adapt, carry forward.

Reports

Evidence before any claim.

Get involved

Run a school, cooperative, NGO, or fund? There's a way in.

Tell me what's 90% there, or what you wish existed by next week. Fixed scope, honest timelines, code you can trust.

Email
contact@mrnm.ai
LinkedIn
linkedin.com/in/m-run-m-ai
WhatsApp
message me
Links
mrnm.link
Card
Save to contacts