# Aravind Srinivas

> ~1993– · AI Researcher, Founder of Perplexity AI
>
> **Recorded contribution:** Perplexity AI; AI search; OpenAI/DeepMind researcher

## How to use this dossier

Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory.

## 1. Historical orientation

Indian computer scientist and entrepreneur Aravind Srinivas completed a Berkeley doctorate, worked in AI research roles, and co-founded Perplexity AI in 2022, becoming CEO. Perplexity popularized an 'answer engine' interface that retrieves current Web documents and synthesizes answers with citations. His contribution is company, product, and systems leadership rather than invention of retrieval-augmented generation or Web search. Perplexity's answer-engine pattern composes retrieval, ranking, language-model synthesis, and citations into one interaction. Its value proposition is not that the model internally knows the latest fact, but that it can locate external documents, condition an answer on them, and expose links a reader can inspect.

## 2. The problem inherited

Conventional search returns links that users must synthesize, while standalone language models can answer fluently from stale parameters without inspectable current evidence.

## 3. The central contribution

Srinivas co-founded and led Perplexity's retrieval-and-generation search product, making source-linked synthesized answers a mainstream interface.

## 4. Reconstruct the mechanism

1. Interpret a question and issue one or more retrieval queries.
2. Fetch and rank candidate documents or passages from current indexes and the Web.
3. Generate an answer conditioned on selected passages while attaching claim-level source links.
4. Let users follow citations and refine queries while monitoring freshness, attribution, and unsupported claims.

## 5. What changed downstream

- Answer engines changed expectations for search interfaces and accelerated competition with conventional search providers.
- The product intensified debates over publisher compensation, crawling, citation accuracy, plagiarism, and concentration of epistemic mediation.
- The product accelerated competition over conversational search and made citation placement, freshness, publisher relationships, and answer verification visible product requirements.

## 6. Attribution, limits, and uncertainty

- Perplexity has multiple co-founders and depends on search indexes, external model providers, publishers, and infrastructure; Srinivas did not invent RAG.
- Citations can be irrelevant or fail to support generated claims, and publisher disputes are material. The registry's approximate birth year is not established here.
- A nearby citation can create unwarranted trust when it does not entail the sentence, retrieval can omit dissenting or paywalled evidence, and synthesized answers can reduce traffic to the publishers they depend on.

## 7. Reconstruction lab

Build a five-document answer engine that must cite every sentence. Add one contradictory and one malicious source; score retrieval relevance separately from answer faithfulness. Score every sentence for citation entailment and source diversity, then compare the concise answer with what a user would learn by opening and reading the cited documents.

## 8. Evidence trail

- [Aravind Srinivas](https://engineering.berkeley.edu/news/2025/03/berkeley-alum-wants-to-make-the-planet-smarter/) — UC Berkeley Engineering
- [Perplexity](https://www.perplexity.ai/about) — Perplexity AI
- [Aravind Srinivas](https://en.wikipedia.org/wiki/Perplexity_AI#Leadership) — Wikipedia contributors

---

*Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*