What does Nikhil Vallishayee actually do?

Three modes, in order of how engagements usually start. **Soundboard** — paid 90-minute sessions for hard, fork-in-the-road decisions. **Audit / Lighthouse** — a 4-6 week deep-dive with a tangible artefact (digital twin, vendor scorecard, board memo, working prototype). **Engagement / Build** — fixed-scope build or milestone-priced phased programme with a tiny team plus an agentic harness. Most engagements walk that ladder: soundboard → audit → build. The thread across all three is the same craft: ThoughtWorks-grade delivery, engineering-org transformation, and agentic AI integration done as engineering, not theatre.

Do you take equity instead of cash?

Sometimes — but rarely as the *primary* compensation. The default is INR cash invoiced milestone by milestone, because that keeps the working relationship clean and the scope honest. Equity makes sense as an *additional* layer on Tier 4 (fractional advisor seat) when the business is pre-revenue and the work compounds for 12+ months. Typical structure: standard advisor grant (0.1-0.5%, 2-year cliff-free vest, single-trigger acceleration on acquihire). I do not take equity-only deals from companies still raising; founders need their cap table for the people building the product full-time, not for advisors.

What's the smallest engagement?

A single Tier 1 soundboard session — 90 minutes, ₹50,000. That's the smallest unit. It's the right shape when you're deciding whether you have a problem at all, or when you've been chewing on an architecture / hiring / vendor decision for two weeks and need a calibrated outsider in the room. I'll read whatever you send ahead (board deck, architecture diagram, code sample, vendor proposal) and come in with a written one-pager after the call. Most soundboards stop after one session; some convert to a Tier 2 audit when the problem turns out to be load-bearing.

What's the longest engagement you've taken?

Four years inside Swiggy as a senior engineering leader, before this advisory practice existed. Inside the current practice (post-2024), the longest single engagement has been a multi-quarter Tier 3 build that started as a 4-week audit and converted into a phased programme. Tier 4 fractional-advisor seats are open-ended and renew quarterly — the longest currently active is 18 months and counting. As a rule I don't take engagements that demand exclusive, full-time presence; the agentic harness is what lets a tiny team move at staffed-team velocity, and that only works fractionally.

Can you start in under a week?

For Tier 1 soundboards — yes, usually within 2-3 working days. Send the situation in an email, we book a slot, and I read into your context the day before. For Tier 2 audits the answer is *probably* — depends on what's already in flight. The audit cadence (4-6 weeks) starts after a one-week scoping pass where we agree on the artefact, the access list, and the success criteria. Tier 3 builds need at least two weeks of lead time because the agentic harness has to be configured against your codebase. NDAs are signed at the engagement-letter stage, not before the first call.

What stays with the team after you leave?

Everything portable. The audit artefact (PDF / Notion / repo). The agentic harness configuration committed to your repo, runnable by any engineer on your team without me in the loop. Architecture decision records (ADRs) for every load-bearing call. A documented operating cadence — what gets reviewed when, by whom, with what evidence. The tiny team I worked alongside, now trained on the harness. Explicitly *not* mine to take away: client IP, customer data, board context, anything covered by the NDA. I run the practice on the discipline that the engagement should make itself unnecessary; if the team needs me to keep things working, the engagement failed.

Yes — mutual NDA at the engagement-letter stage, before any privileged material moves. The first 30-minute exploratory call is informally confidential by default; nothing from it is published, blogged, or shared with another client without explicit written consent. For Tier 2+ I'll counter-sign your standard mutual NDA if it's reasonable; if your template has unusual terms (perpetual non-compete, IP-grab clauses, jurisdiction-only-in-X) I'll redline. I do not sign one-way NDAs that bind me but not you — both sides exchange privileged context in this work, and the protection has to be symmetric.

Can I see a sample audit deliverable?

Two case studies are public on this site — the Navi neo-bank greenfield platform and the Swiggy food-delivery org-shape transformation — written at the depth of an actual audit memo. They're at `/work/neobank-greenfield-platform` and `/work/food-delivery-org-shape`. For redacted private samples (a board memo, a vendor scorecard, a digital-twin walkthrough) ask in the first call and I'll send what's appropriate to your situation, under NDA. The shape is consistent across audits: a one-page executive summary, a load-bearing-decisions list, a risk register, and an artefact (twin / scorecard / prototype) sized to the question.

What does Tier 2 actually cost?

₹15 lakh – ₹50 lakh for a 4-6 week audit, billed as 50% on engagement-letter signature, 25% at the midpoint review, 25% on artefact handover. The variance is driven by depth (one team vs. a portfolio), travel (Bengaluru-only vs. on-site weeks elsewhere), and the artefact (a board memo is lighter than a working prototype). For pre-Series-B startups I'll usually quote at the lower end and structure milestones around your runway. USD-equivalent benchmarks on request for international clients — past Tier 2 work has been quoted at $25K-$80K depending on scope.

Do you work with pre-product startups?

Yes — and it's where the Tier 2 *Lighthouse* shape was invented. Pre-product means there's no system to audit yet, so the audit becomes a forward-looking artefact: a working prototype that proves the riskiest assumption, or a digital twin of the business model that lets you test pricing / channel / unit-economics before writing production code. The SF crypto-underwriting MVP (a 21-day Tier 2) is one such engagement — pre-product when we started, in market within three weeks. Pre-product engagements work best when the founder is technical enough to keep the cockpit and is hiring an engineer-zero in parallel; my job is to make their first hire massively more effective, not to be the first hire.

What's your background?

Twenty years in software, anchored by ThoughtWorks for the deep delivery craft and four years inside Swiggy as a senior engineering leader during the post-pandemic scale wave. In between: an end-to-end neo-bank build for Sachin Bansal at Navi (greenfield platform, ~170 services, regulator-grade); a real-time matchmaking engine for a Bay-Area billionaire; a leadership-development platform with Minal Mehta; a 21-day MVP for an SF crypto CTO. I founded Breathworks Universal Systems in 2026 to house the agentic-harness IP that powers the current advisory practice. Bengaluru-based, vendor-neutral, works in IST + PST overlap, English and Hindi.

Where are you based and what timezones do you cover?

Bengaluru, India. The advisory practice is globally vendor-neutral. Best timezone overlap is IST + PST — early morning India / late evening US Pacific — which is the daily window when I run live sessions with both Indian and Bay-Area clients. London / Singapore / Sydney clients work fine on a same-day basis. I travel for on-site weeks during Tier 2 and Tier 3 engagements when it materially helps the work; everything else is remote with a high-quality recording / artefact discipline so the team can replay context async.

Food-Delivery Marketplace — Reshaping the Engineering Organisation

Case study, names changed for confidentiality. Synthesised from a four-year engagement with a top-three Indian food-delivery marketplace.

The brief

The marketplace had the hardest scaling problem most Indian product companies face: hire fast enough to keep up with the platform, but hire fast enough that the quality bar drifts every quarter. Their answer was an internal associate-engineer programme — a structured nine-month onboarding for engineers in their first two years.

The leadership ask was simple in the framing and hard in the doing: how do we build the programme so the engineers who graduate ship production-grade work, not "ready for code review" work?

The capability axes I held

4 yearsembedded engagement

200+engineers shaped

9 monthcohort cycle

I was the technology subject-matter expert facilitating the programme. The role was deliberately ambiguous — closer to a curriculum-and-coaching hybrid than a teaching role.

Three threads ran through it:

Curriculum that matched the codebase. Generic courseware loses to the actual repo within a week. We rebuilt the curriculum around real services and real incident reports — the engineers learnt distributed-systems thinking from the systems they were about to touch.
Pairing as the primary instrument. Lectures don't change behaviour; pairing does. Every cohort engineer paired weekly with a senior reviewer, and the senior was held accountable for the cohort engineer's growth, not just their PR throughput.
Recovery loops, not gates. The traditional model is graduate-or-fail. We built recovery loops instead — an engineer who missed a milestone went into a remediation pairing rotation rather than a "PIP" track. The dignity preserved was material; the retention number doubled.

What surprised me

That one constraint changed the org's hiring posture more than any training programme could have. It said, out loud, that quality was a manager's responsibility too — not just the cohort facilitator's.

How I measured it

The honest answer: imperfectly. The defensible numbers were retention (up significantly) and time-to-first-production-incident-handled (down significantly). The harder-to-defend number was the qualitative one — peer reviews of the cohort graduates' code six months out, by senior engineers who didn't know which cohort the author came from. Those reviews were the most important signal we had.

What I'd reuse

Curriculum on the actual codebase. Day one.
Senior accountability for cohort growth. Not optional.
Recovery loops over graduate-or-fail gates. Always.
Receiving-manager commitments in writing. The single highest-leverage intervention for any cohort-style hiring programme.

Why it mattered

Companies at this scale don't fail from architecture. They fail from hiring drift — from the slow erosion of the bar that, two years later, makes the codebase unrecognisable to the people who built it. Cohort work is one of the few interventions that meets that drift head-on.