Solutions Engineer · Production LLM Applications
Recently shipped an LLM medical records solution to production at a Fortune 500 U.S. insurer as sole technical lead. Open to Solutions Architect, Forward Deployed Engineer, and Solutions Engineer roles at LLM-first companies. Based in India, open to relocation anywhere in the world.
A quick intro
I turn enterprise AI proofs-of-concept into production systems.
Five-plus years as a Solutions Engineer for Fortune-100 insurance and Tier-1 banking customers, as the sole technical owner on every account I run, no supporting engineer. I own the technical motion end to end, from the first pre-sales POC through deployment, and I build for reuse, so the systems I ship keep working long after handoff.
Rahul Nanwani
01 · Case study
Convinced a Fortune-100 U.S. life insurer to build less than they asked for.
01
The customer came in with business questions, not a technical spec. I worked through several iterations with them to translate what they needed into the actual document classes required.
02
They were leaning toward split classification, dividing each document into sections and classifying every split. I made the case for simple classification, one class per document, instead, on accuracy and on cost.
03
Built the classification workflow, consolidating 40-50 page claims documents into structured, analyst-ready summaries. Validated it against the customer's ground truth, as sole technical owner.
04
Shipped to production at around 30,000 pages a year. The simpler model matched the complex one on accuracy, at a lower cost to run, and the customer is happy with the result.
05
They now ask for me by name when new work comes up on the account, not whoever's assigned.
02 · Proof
03 · Delivery
Fortune 500 U.S. insurer
Shipped a medical records extraction solution to production as sole technical lead for the period I owned it end to end, processing around 4.5 million pages per year. The end-to-end design was adapted from an existing use case at the same customer; I carried it from development through go-live and now own it in production.
~4.5M pages/yrTier-1 UK insurer
Shipped a postal claims-intake solution to UAT completion at 96.2% field automation accuracy across 70 fields and four deep-learning models. Designed the solution architecture, trained a new engineer mid-project, and recovered an aggressive timeline after a parallel production incident.
96.2% accuracy across 70 fieldsglobal insurance carrier
Led a competitive technical evaluation to automate extraction from Japanese handwritten medical certificates. The pipeline handles era-based date conversion (Japanese imperial calendar to Gregorian), hanko stamp recognition, signatures, and circled checkbox selections that the text-only baseline could not address. We were the preferred vendor on the technical evaluation; the customer ultimately chose to build the solution in-house.
$1.3M TCV / $433K ARR · technical evaluation, preferred vendorBig Four Australian bank
Built a fraud-detection proof of concept that flags tampered paystubs embedded inside larger, multi-document packets, first by integrating a third-party document-fraud-detection engine into the extraction pipeline, then by replacing it with an LLM-based tool of my own: a detailed fraud-detection prompt run directly against the uploaded document on our production LLM platform, that outperforms the third-party engine.
Technical POC · own LLM tool outperforms third-party engine04 · Outside work
A complete library for blackjack rounds: multi-player tables, configurable multi-deck shoes, the full action set (hit, stand, double-down, split, surrender), and configurable dealer rules. Versioned releases on PyPI with a test suite, CI, and documentation on Read the Docs. Latest release: v5.0.0 (March 2026).
05 · Method
I build reusable tools, not one-off scripts. Two of mine are now standard across the org.
06 · Path here
Apr 2024 – present
Solutions Engineer II
Production LLM applications for Fortune-100 insurance and banking, from pre-sales POC through deployment.
Oct 2022 – Mar 2024
Solutions Engineer I
Shipped claims-indexing and extraction solutions to production across insurance and banking.
Jul 2022 – Oct 2022
Data Operations Lead
Led a team of associates and cut project ramp time by 3+ weeks with new data-preparation strategies.
Aug 2021 – Jun 2022
Data Operations Associate
Contributed to 10+ client use cases and 5+ pre-sales POCs across banking, insurance, and mortgage.
2024 – 2025
M.Sc. Machine Learning & AI
Liverpool John Moores University, UK
Thesis: Explainable AI for Image Classification.
2023 – 2024
Executive PGP, Machine Learning & AI
IIIT Bangalore, India
2017 – 2021
B.Tech, Computer Science & Engineering
MIT ADT University, India
07 · Right now