Mrunmay Phanse — AI Engineer

Mrunmay Phanse is an AI engineer who builds retrieval-augmented generation systems and the named author of RAG Explained Better. The facts below state who he is from what is already published, how to verify him through outside profiles, how work on the site is done, and how to reach him, without inventing employers, years of experience, or project counts that have not been written down.

Who is Mrunmay Phanse?

Mrunmay Phanse is an AI engineer who builds retrieval-augmented generation systems. That is the role claim asserted here because it is the claim already stated on About RAG Explained Better: one accountable name behind every byline, not an anonymous content mill. The author entity exists so a sourced number or a worked example can be weighed against a checkable person rather than against a brand voice with no outside record.

A fuller biography and a real headshot are still owed. Until they land, employers, years of experience, client names, and project counts are not invented to fill the gap. An author entity that fabricates credentials destroys the trust advantage a named engineer is meant to create. The honest publication gap is the gain: verifiable identity without a fluff CV.

Site funding and the Weaviate placement rule live on the about page, not here. The person page states identity; the about page states money. Mixing those jobs is how an author bio turns into a sponsorship speech.

What can be stated without invention is narrow on purpose. The role is AI engineer. The subject is retrieval-augmented generation. The publication is RAG Explained Better. The outside profiles are the four URLs already carried in schema. Everything else that would normally pad an author page (employers, degree years, conference talks, client logos) waits for a real biography rather than a fabricated one.

What does Mrunmay Phanse write on RAG Explained Better?

Mrunmay Phanse writes every page on RAG Explained Better: technical writing on how to build retrieval-augmented generation, measure it, and understand why it fails. Each page carries his byline, so a pinned library version, a failure diagnosis, or a comparison grid is a person putting a name to the claim. The site covers retrieval mechanics, chunking, evaluation, architectures, infrastructure, and failure modes as a course a practitioner can work through in order.

The writing is held to current primary sources and to the rule that every number is sourced or omitted. How the site chooses sources and checks claims is summarised on About RAG Explained Better; the deeper quality bar for research, citation, as-of dating, and measurement pages is on the methodology page. The author entity is the person behind the byline, not a second copy of those policy pages.

Readers who want the sponsorship disclosure should open the about page. Readers who want the benchmark setup, hardware pins, and cite rules should open methodology. Readers who want to know who is accountable for a wrong number should stay with the author entity and the contact route below.

Topics under the byline include why retrieval returns the wrong chunk, how hybrid search fuses dense and sparse signals, how faithfulness differs from answer relevance, and how architectures from naive RAG to agentic RAG change failure modes. Each of those subjects has its own mechanism page; the author page does not redefine them. The glossary is the index of terms; the hubs are the course chapters; the author page is the accountable name on the syllabus.

How can you verify Mrunmay Phanse?

Mrunmay Phanse is checkable through four outside profiles that match the site’s Person schema sameAs record, so the person behind the claims can be weighed, not just named. The URLs are github.com/mrunmayphanse for code, linkedin.com/in/mrunmay-phanse-8a364b56 for the professional record, mrunmay.dev for the personal site, and dev.to/mrunmayphanse for writing. Those four strings are the same values carried in the theme’s author schema.

A name alone is not an author entity. Opening the profiles is how a reader corroborates that the byline points at a real person with a public trail. No follower counts, star counts, or employment dates are invented from those profiles; the verification job is identity, not a scraped resume. If a profile URL changes, the schema and the verify list update together so the sameAs set stays consistent site-wide.

External profile pages are named in prose and are not linked as outbound anchors from article body copy, because the site’s link rule keeps PageRank inside the RAG Explained Better network while still giving a reader a string they can paste. Schema markup remains the machine-readable sameAs set for search systems that consume it.

Verification does not require trusting a vendor press kit. The four URLs are ordinary public profiles. A reader who cannot open them should treat the byline as unchecked until the schema and the named strings are repaired. That failure mode is rare, and naming it is still better than inventing a private email that was never published.

How does Mrunmay Phanse work on this site?

Mrunmay Phanse builds pages from current primary sources, with every number sourced and never invented: the same method every substantive page on the site is held to. Author authority here rests on a stated method, not on an uncheckable resume. Vendor documentation, papers, and maintained repositories are preferred over secondary summaries, and volatile claims carry an as-of date in the prose itself.

The benchmark setup, the versions, the harder bar for site-run measurement pages, and the cite rules are spelled out on how we test, measure and cite. The author page does not restate that entire quality bar; it points there so a reader who trusts a byline can audit the method behind it. When a page reports a score the site computed, the pins required for reproduction live on methodology and on the measurement page itself.

Work that only cites someone else’s published benchmark still names the maintainer and the as-of date. Work that invents a hardware SKU or a latency the site never measured does not ship. That discipline is why the author entity can stay short on credentials and long on method routing.

AI-assisted drafting tools may be used as writing aids the way a spellchecker is used, but AI output is never treated as evidence for a number, a capability, or a limitation. The bylined author remains accountable for every claim that publishes. The fuller statement of that rule sits on the methodology page beside the cite and as-of rules, so the author page does not become a second editorial-policy essay.

How do you contact Mrunmay Phanse?

Reach Mrunmay Phanse through any of the four profiles named in the verify section; GitHub at github.com/mrunmayphanse is the fastest path for a factual correction. When a number has drifted, a source has moved, a version has been renamed, or a claim no longer holds, corrections are made on the page itself, with a visible note when the change is material, not argued only in a comment thread.

A named author page that cannot be reached is not an author entity. Contact here is for factual fixes and identity checks, not for sales demos or undisclosed guest posts. Sponsorship questions and the placement-not-inflation rule belong on About RAG Explained Better; method and measurement questions belong on the methodology page; person and corrections belong here.

If a correction changes a number that other pages repeat, the same value is updated everywhere it appears so the site keeps one fact, one value. Silent stealth-edits of wrong figures are not the policy. Reporting an error is how the byline stays meaningful after publish.

Email addresses and phone numbers that have not been published as contact facts are not invented here. The four profile URLs are the published contact surface. Readers who prefer LinkedIn or the personal site can use those strings; readers who want a tracked issue for a factual fix should prefer GitHub because the correction trail stays visible next to the code identity.

Guest-post pitches, undisclosed affiliate offers, and requests to alter a score without new evidence are declined. The author entity exists to make accountability checkable, not to turn the byline into a soft inbox for paid placement. Commercial disclosure for the site as a whole remains on the about page under the Weaviate funding sections.

Production experience claims, publication counts, and conference talk lists that appear in the topical-map meta description are not restated as body facts until a real biography supplies them. Meta marketing language does not license invented credentials in the body copy.

If you opened this URL from a byline, the next useful stops are About RAG Explained Better for funding and placement, and how we test, measure and cite for the quality bar behind every sourced number on the site.

Mrunmay Phanse is an AI engineer who builds RAG systems and the named author of RAG Explained Better. The only role claim asserted is the one already published on the about page; a fuller biography and headshot remain owed rather than invented. Four sameAs profile URLs make the person checkable, methodology carries the quality bar, and factual corrections land on the page through those profiles.