Who owns the renewal?
Srikanth Gopalakrishnan · RevOps systems architect
If sales won't
use it Tuesday,
it stays draft.
I design and ship the operating system: HubSpot architecture, a Snowflake / dbt platform, and agents that put the next action in Slack — with evals and metrics a board can reproduce.
The questions a board already has. I answer them with a system.
Can the board number be reproduced?
Is the agent still a demo?
Will anyone miss this Tuesday?
What I do
Design the system. Then ship it.
Four ways I show up. Commercial judgement first: choose the right system. Then leave something that still runs.
- 01
Solution design
Architecture before build so the system survives the next hire. CRM truth, stage gates, and a sequence sales will actually run.
- 02
Revenue platform
Snowflake / dbt / Airflow plus HubSpot governance that matches the board. Reproducible ARR and hygiene — not another Excel fight.
- 03
AI agents with receipts
Slack and WhatsApp agents with guardrails and eval suites. Draft-to-live, not a demo that dies in week two.
- 04
Fractional systems architect
Roadmap, gates, and operating rhythm. I sit with your team 1–2 days a week until the system runs without heroics.
The front door
Peak Check
A diagnostic call and a one-page map. Choose the right system to ship — before another tool, hire, or agent.
60–90 minutes + a written 1-pager
One working session. I write the page after. Optional short debrief if you want to walk it together.
Who it’s forRevOps, Sales Ops, and GTM leads who need commercial judgement, not another demo.
When to hireThe board number is political, renewals surprise you, or an agent is still stuck in draft.
What we go through
- 01
CRM truth
Owner, stage, property hygiene. What sales actually trusts versus what the CRM pretends.
- 02
Renewal risk
What is invisible, unowned, or still showing up as a surprise.
- 03
Board metrics
Whether ARR / NRR can be reproduced — or still depends on who built the tab.
- 04
Automation readiness
What is worth shipping, and whether anyone will sponsor it on Tuesday.
- 05
Agent eval readiness
Whether an agent should stay in draft, and what the judge harness must catch first.
Deliverable
1-page “what’s broken / ship next”
Prioritized and named. Short enough that a sponsor can act. Optional short debrief — same page, no second deck.
What happens next
Systems Sprint
A scoped 2–6 week build if the page names one thing worth shipping.
Fractional seat
1–2 days a week if the roadmap is longer than a sprint and someone has to own the gates.
Stop here
You keep the page. No retainer by default. I will say so if a sprint is the wrong next step.
More value comes from choosing the right system to ship, not from shipping more automations.
Fit and scope are decided on the call. No menu of retainers.
Book a Peak Check
Prefer email? hello@thepeaktree.com
If we keep going
A sprint, or a seat. Not a retainer by default.
The Peak Check is the front door. These are how we continue — only if the page names work worth doing.
2–6 weeks
Systems Sprints
A scoped build with a named sponsor. Hygiene scoring, renewal-risk nudgers, board ARR on Snowflake, CRM governance, or an agent with an eval harness.
Who it’s for
Teams that already know the messy bit and have someone who will use the thing on Tuesday.
When to hire
You need it in production this quarter — not a roadmap workshop.
1–2 days / week
Fractional RevOps Systems Architect
Ongoing roadmap ownership, draft-to-live gates, and mentoring so the operating system doesn’t depend on one hero.
Who it’s for
B2B teams (or a boutique agency) that need a senior builder in the stack, not a slide deck.
When to hire
The work is longer than a sprint and someone has to own gates, metrics, and the next hire’s inheritance.
How engagements run
I prove the system, then I leave you owning it. Fractional exists when the roadmap still needs an architect — not as the default.
- 01
Design.
Architecture, success criteria, and draft-to-live gates before a line of automation. Named sponsor. Read-only until the criteria are real.
- 02
Ship.
Build on HubSpot, Snowflake / dbt, and agents with evals. Draft first. Live only when someone has used it and will miss it.
- 03
Hand over
Docs, Metrics Book–style lineage, and a named owner in the business. They run the system. I do not stay on a retainer by default.
Selected work
War stories, not case-study theatre.
EdTech B2B systems, a named enablement build for Avon Seals, and methods you can defend. I sanitize numbers. I do not sanitize the standard.
Live
Avon Seals · Chennai · assessment app
Baseline before the room
Management Development Assessments for a leadership program I delivered. The artifact is an interactive app — still live — not a slide deck after the fact.
Read the story →
~$125k
EdTech B2B · orphan ACV
Renewals no rep could see
A Slack nudger fused live CRM and call notes so renewals without an owner stopped being invisible.
14/15
EdTech B2B · LLM-as-judge
The AI SDR had to pass a test first
WhatsApp outreach with deterministic guardrails and a judge harness before anyone talked to a customer.
How I work
“I would rather ship five automations that get used than fifteen that look good in a demo.”
- Draft
- Eval / sponsor
- Live
01
Draft-to-live
Nothing ships because a ticket closed. An automation goes live when someone has used it in draft and will miss it if it disappears.
02
Read-only by default
Agents observe and recommend before they write back to HubSpot. Privilege is earned after the eval suite and a named owner say so.
03
Named sponsor
Every build has a human in the business who owns usage. If nobody will defend it on Tuesday morning, it does not ship.
What I put in production
HubSpot governance · Snowflake / dbt / Airflow · LLM-as-judge evals · Slack & WhatsApp agents
Background
Depth behind the systems
Agents and evals hold up because I used to ship models. This is not a second offer — it is why draft-to-live has teeth.
- ML / DL
- RAG
- ML CI/CD
- Computer vision
- Forecasting
- Python
01 ML / DL · Python
Ericsson: 57% cloud-cost reduction with utilization criteria and written diagnostics — not one-off scripts.
02 ML CI/CD · RAG
Telecom knowledge systems with RCA loops and answer-quality review before anyone trusted the retrieval.
03 Computer vision
Gramener: liposome counting from 540 hours to 1 hour, with a success criterion you could defend.
04 Forecasting
Ongil: stock-out risk models that cut losses ~150k SGD versus the operational baseline.
05 Teaching
DS practitioners at UpGrad and PES, plus Ericsson’s ALTitude program — judgment standards, not tool demos.
About
Srikanth Gopalakrishnan
I build the revenue operating system — then I leave you with something that runs. Peak Tree is the practice. I am the person who shows up.
One-person. Not a faceless agency, not HubSpot-only, not an AI strategy deck. Commercial judgement first: choose the right system to ship. Then I implement it.
Before this I led data science at Ericsson and shipped ML for enterprise teams. Rubrics, baselines, written forensics. The surface changed. The judgment did not.
India · APAC, Japan, LATAM, US. LinkedIn