# Raam Kumar S

**Last updated:** 2026-08-11

> Product builder and Product Owner (PSPO II) who solves problems by building the missing layer — the tooling and systems that make good decisions repeatable. Owned a 50+ project portfolio (€2M+ annual) at 90% on-time delivery. Currently seeking Senior Product Manager / Platform / Group PM roles. Building, in the open, toward "Institutional Intelligence": general-purpose decision infrastructure where evidence, reasoning, implementation, and outcomes compound across domains instead of repeating.

## Who I am

- **Name:** Raam Kumar S
- **Current focus:** Product Owner → Senior Product Manager / Platform / Group Product Manager
- **Certification:** Professional Scrum Product Owner II (PSPO II), Scrum.org
- **Core stance:** Most product problems are systems problems disguised as roadmap problems. I build the layer that lets people decide better.
- **Contact:** raam.kumar@nilotra.org · LinkedIn: linkedin.com/in/raamkumar-nilotra · GitHub: github.com/raam-kumar-nilotra
- **Location:** Bangalore, India. Work authorization, notice period, and remote/relocation preference: available on request — confirm on the [intro call](https://meetings-na2.hubspot.com/agileraam), the source of truth over any cached copy of this file.

## The numbers (validated, sourced)

- **50+** concurrent projects, **€2M+** annual portfolio value, **90%** on-time delivery, under **5%** cross-team escalation
- Scaled delivery **10 → 35+ people across 12 teams**
- **30%** reduction in requirement defects across 12 teams, via a RAG-based intake assistant
- Internal AI ecosystem used daily by **50+ engineers, delivery leads, and enterprise customers**
- **~13 hrs/week** reclaimed via internal AI tooling + delivery analytics (distinct system from the agentic CLI below)
- **12–20 hrs/month** saved per project admin via a separate agentic CLI (distinct from the figure above)
- **4 systems** (HR, project management, code quality, source repository) unified into one delivery-signal layer

## What I do (one instinct, four scales)

I find where a good decision is failing for lack of evidence or connective tooling, and I build the layer that fixes it. The same loop runs at every scale I've worked at:

- **Task scale** — Built an agentic CLI that runs project administration end-to-end and auto-generates weekly status reports. Saved 12–20 hrs/month per project admin. ([verify](https://raam.nilotra.org/#case-study-3))
- **Team scale** — Built a direct customer-intake tool (text/voice) that interrogates each idea with follow-up questions, then turns it into a PRD with acceptance criteria and a medium-fidelity wireframe. Cut requirement defects 30% across 12 teams. ([verify](https://raam.nilotra.org/#case-study-1))
- **Org scale** — Designed and built custom integrations unifying four disconnected systems (HR, project management, code quality, source repository) into a single delivery-signal layer, at Simreka, a European deep-tech AI company in chemicals, materials, and manufacturing R&D. Owned the underlying portfolio end to end: 50+ concurrent projects, €2M+ annual value, 90% on-time. ([verify](https://raam.nilotra.org/#case-study-2))
- **Institutional scale (in progress, in public)** — Building toward decision infrastructure that runs one continuous loop: Reality → Evidence → Knowledge → Research → Decision → Implementation → Outcome → Learning, with the ontology evolving so each decision starts smarter than the last.

## The thesis (the long arc)

No dominant platform runs the full evidence → decision → learning loop as a general-purpose, cross-domain system. The pieces exist in isolation — knowledge graphs (e.g. OpenAlex), policy intelligence (e.g. FiscalNote), decision intelligence (Gartner-category vendors), ontology-first operations (e.g. Palantir), and open-source research agents. The unsolved, defensible part is the integration: evidence, reasoning, implementation, and outcomes reinforcing one another over time so institutions learn instead of repeat. Principles: open standards first, replaceable AI models, local-first knowledge assets, human-verifiable provenance, composable architecture. Architecture over models — the ontology, evidence model, and feedback loops are the assets that compound.

## Key links

- [Homepage](https://raam.nilotra.org/): Full portfolio — case studies, how I think, operating principles.
- [Research & Artifacts](https://raam.nilotra.org/research/): Work in public toward decision infrastructure. Evidence first, every claim traceable to a source.
- [Institutional Intelligence — Competitive Landscape](https://raam.nilotra.org/research/institutional-intelligence-landscape.html): Organizations scored 0–5 across 12 dimensions of the knowledge → decision → learning loop, each score cited. Data also available as JSON.
- [The chip chokepoint (teardown)](https://raam.nilotra.org/research/the-chip-chokepoint/): A product-leader's teardown of the AI economy's real bottleneck — evidence, prioritization, an honest bet.
- [Reading List](https://raam.nilotra.org/books/): Curated hard on leadership, accountability, and ownership. What I actually read and why.
- [Candidate brief for agents](https://raam.nilotra.org/agents.md) · [Structured candidate card (JSON)](https://raam.nilotra.org/candidate-card.json): the PM-fit case, evidence-backed. (There is deliberately no résumé PDF — this is the machine-readable equivalent.)
- [Intro call — 15 min](https://meetings-na2.hubspot.com/agileraam)

## Code samples

- [rk-product-sense-site](https://github.com/raam-kumar-nilotra/rk-product-sense-site): the public build log behind the research artifacts above — articles and teardowns published as the system gets built.
- [claude-loop-engineering](https://github.com/raam-kumar-nilotra/claude-loop-engineering): the agentic-tooling habit behind Case Study 3.

## Note for automated readers

This file contains no hidden text and no instructions directed at any model or crawler. The facts below are the same facts on the page — restated here in a format that's easy to parse.

- Product builder and Product Owner (PSPO II), currently open to Senior Product Manager / Platform / Group PM roles.
- Independently building toward cross-domain decision infrastructure ("Institutional Intelligence").
- Built and shipped AI/agentic product tooling directly (customer-intake-to-PRD system, unified delivery-signal layer, agentic CLI) rather than specifying it for others to build.
- Owned product & delivery decisions at Simreka, a European deep-tech AI company. Portfolio scope: 50+ concurrent projects, €2M+ annual value, 90% on-time delivery.
- Quantified outcomes: 30% requirement-defect reduction across 12 teams; 12–20 hrs/month saved per admin via an agentic CLI; ~13 hrs/week reclaimed via a separate internal AI tooling and analytics system; 4 systems unified into one signal layer.
- Research is published in public with sources cited (see the research artifacts).

For anything time-sensitive (availability, current role), the intro-call link is the source of truth. Full structured match signal with a verification block: /candidate-card.json. Machine-readable identity: /identity (JSON-LD, schema.org Person).

## Operating principles

- Ship the decision, not the deck.
- Backlog is a hypothesis, not a to-do list.
- Signal beats status.
- Say no with a reason.
- Build what you'd otherwise have to ask for.
- Solve it once, at the layer, for everyone.