Skip to content
YC Root AccessYC Root Access

Salient: The Fintech Startup Processing $1B+ in Loans with AI

Ari Malik and Mukund Tibrewala started Salient after seeing firsthand, during Ari's time at Tesla, how expensive and outdated loan servicing really was. What began as a side project—automating outbound calls with voice AI—quickly evolved into something bigger: a fully integrated, AI-powered loan servicing platform built for both fast-moving non-bank lenders and some of the largest banks in the U.S. In this conversation, Salient's co-founders share how a single cold email and a Steve Jobs–voiced demo landed Salient its first major customer. They talk about moving across the state to sit next to that customer until they were live, scaling from hundreds to hundreds of thousands of daily calls, and navigating one of the most complex regulatory landscapes in tech. Salient now processes billions in loans, serves millions of borrowers, and just raised a $60 million Series A led by Andreessen Horowitz—all with a team of 10 engineers. This is the story of how they did it. Learn more about Salient at https://www.trysalient.com. Apply to Y Combinator: https://ycombinator.com/apply Chapters: 00:22 - What Salient Does 01:03 - Early Beginnings at Tesla 01:26 - Leveraging AI for Loan Management 01:51 - Scaling with Open Source Models 02:15 - The Impact of LLAMA 2 02:39 - The Magic Demo and First Big Customer 03:12 - Cold Emails and Westlake Financial 05:02 - Relentless Customer Focus 05:35 - Forward Deployed Engineer Playbook 06:19 - Scaling with a Small Team 07:13 - Hiring for Growth 07:47 - Challenges in Scaling Voice AI 08:24 - Navigating Regulations and Compliance 08:56 - Future Vision for Salient

Diana HuhostAri MalikguestMukund Tibrewalaguest
Jul 27, 20259mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Salient scales AI loan servicing from demos to regulated production systems

  1. Salient provides an AI loan servicing platform for consumer lenders, expanding from auto loans into broader products like credit cards and mortgages.
  2. The founders trace the idea to Tesla, where they saw loan management costs remain high even for low-risk borrowers, signaling large automation potential.
  3. They used closed-source models to create compelling early demos, then relied on open-source models (notably Llama 2) and tooling like vLLM to scale economically from small pilots to massive call volumes.
  4. Early growth came from relentless outbound sales (hundreds of cold emails daily) and an extreme “forward deployed” approach, including relocating near a first major customer to get fully live.
  5. Scaling required more than model quality: Salient built compliance, dialing legality, and operational safeguards (e.g., TCPA rules, bankruptcy and legal-representation protections) to run voice AI safely at high volume.

IDEAS WORTH REMEMBERING

5 ideas

A great demo can unlock the first enterprise door, but scaling requires a different stack.

Salient used GPT-era capabilities to create a convincing “magic demo,” then switched to fine-tunable open-source models and efficient inference (vLLM) to make high-volume production economics work.

Open-source models can be a decisive scaling lever when cost per interaction matters.

They describe Llama 2 as closing the gap between impressive pilots and sustainable operations, enabling expansion from ~100 calls/day to hundreds of thousands of calls/day at far lower cost.

Landing big customers can come from brute-force outbound paired with extreme follow-through.

The team sent ~500 cold emails/day and, after Westlake responded, moved near the customer and spent ~18 months getting the deployment live—out-executing larger vendors.

“Forward deployed engineers” compress time-to-value in regulated enterprise workflows.

Engineers owned major customers directly—mapping workflows, deploying into production, and speaking with executives (CFO/VP of servicing)—rather than handing off after delivery.

Voice AI success in lending is constrained by compliance and operational safety, not only model quality.

Beyond low latency and natural interruption handling, they emphasize dialing only when legally allowed, managing thousands of concurrent dials, and embedding protections tied to regulations like TCPA.

WORDS WORTH SAVING

5 quotes

Salient is an AI loan servicing platform for consumer lenders.

— Ari Malik

We could actually automate a lot of the stuff using AI.

— Ari Malik

The advent of open source models... really, really helped us scale from, like, a few hundred dollars per day to over $100,000 per day.

— Mukund Tibrewala

We cold emailed every single auto lender under the sun... Like, 500 a day.

— Ari Malik

Our goal is to be the system of record for every loan in America.

— Ari Malik

AI loan servicing for consumer lendersOrigin story at Tesla and cost-to-serve problemClosed-source demos vs open-source scalingLlama 2 and vLLM for cost/performanceCold-email outbound and landing Westlake FinancialForward-deployed engineer / Palantir-style deploymentsVoice AI at scale: latency, interruption, compliancePCI and regulatory safeguards (TCPA, state/federal rules)Small-team scaling and high-agency hiringVision: system of record for every loan in America

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome