RN Rishabh Natarajan / SA · Builder · Producer

Applied AI · Solutions Architecture · Build log

I build calibrated ML systems and translate them for the people buying.

Solutions Architect, Amazon Ads (L5). Currently building Transferithm — a source-graded transfer-rumour tracker with a public accuracy record, launching around the World Cup. This is where the architecture decisions, build logs, and longer case studies live.

Now writing  Driving an LLM pipeline to ~$0 · Jersey City, NJ · ET · rishabh@rishabhnatarajan.com

Recent writing

What I've been working through

All writing
  • 2026 · 06 · 21

    Driving an LLM pipeline to ~$0 without losing accuracy

    Cutting two high-volume LLM stages to roughly $0 by treating cost as an eval problem — build the measurement first, prove a free open-weight model matches (and on extraction beats) the paid one, then ship an eval-gated router. Plus the honest sequel: the $0 swap that quietly froze ingestion for a week, and what the eval never measured.

    case study · 16 min
  • 2026 · 06 · 16

    Linking World Cup performance to transfer rumours — honestly

    The marketing question was 'does a great World Cup actually move a player's transfer market?' The naive answer — infer the link ourselves — is unprovable from observational data and would break the product's honesty firewall. The honest move was to stop inferring and instead detect where a journalist already stated the link, attribute it, and measure whether a move follows. A measurement-first pass showed a real-but-small signal (n=9), which is exactly why the next step is to capture labelled data, not re-measure a tiny sample.

    case study · 11 min

Featured projects

What I'm building

All projects
Transferithm Active

A single, source-graded feed for following your club's transfer window — with a public track record.

A multi-source data pipeline that ingests transfer reports from dozens of football outlets, uses an LLM to extract each player-to-club link, grades it by source reliability and corroboration, and publishes a verifiable track record. Eval-first; the LLM layer runs on an open-weight model at roughly $0/mo.

typescriptnextjssupabasepostgreslangchainlangfusesportmonksvercel

Now

Currently focused on

About me

Shipping

Transferithm — World Cup 2026 launch: a source-graded transfer-rumour feed with a public track record, LLM pipeline running eval-gated at ~$0/mo.

Reading

LLM evaluation and calibration; how public forecasters publish and grade their own accuracy.

Practising

Turning build decisions into case studies I can defend in an interview.