---
title: "Nikhil Shetty - Senior Data Scientist in Bengaluru | Tal"
description: "Meet Nikhil Shetty, Production AI Systems for BFSI | Agentic RAG, Document AI, LLM Evals, vLLM/K8s | Building AI Systems Banks Can Trust based in Bengaluru. AI stack includes anthropic apis, codex, and hermes agent."
image: "https://tal.club/meet/nikhil-raj-shetty/share-image"
---

# Nikhil Shetty on Tal

## Headline

Production AI Systems for BFSI \| Agentic RAG, Document AI, LLM Evals, vLLM/K8s \| Building AI Systems Banks Can Trust

## About

I build AI systems for financial services, mostly around GenAI, document intelligence, RAG, and production ML APIs\. The thing I’ve learnt clearly: AI in banking is very different from AI in a demo\. A demo can look great on a clean PDF and one happy path\. Production has to deal with bad OCR, missing fields, latency pressure, hallucination risk, audit requirements, and business teams who need answers they can actually trust\. At Arya\.ai, I’ve worked on BFSI AI systems across KYC, bank statement analysis, credit memo automation, and enterprise GenAI products\. A big part of my work has been around production document intelligence APIs for Axis Bank, along with scalable GenAI/RAG systems using FastAPI, Docker, Kubernetes, vLLM, and PagedAttention\. What I enjoy most is the mix of engineering and business context understanding the workflow, knowing where the risk is, and building systems that are useful beyond the demo\. Long term, I want to grow at the intersection of AI, financial services, and business leadership\. Building AI systems banks can trust\.

## Location

Bengaluru

## AI stack

anthropic apis

codex

hermes agent

hugging face

langchain

langgraph

## Work experience

Senior Data Scientist at Arya\.ai \- An Aurionpro Company, Dec 2025 \- Present · 10 mos, On\-site, • Built an Agentic RAG pipeline for credit memos using hybrid retrieval and Cross\-Encoder re\-ranking; integrated web\-search tools to cut report turnaround from 3 weeks to 3 days \(85% faster\)\. • Owned technical strategy for 3 enterprise GenAI products; designed a multi\-tenant inference architecture on K8s/vLLM, sustaining 99\.9% uptime and improving operational throughput by 20%\. • Benchmarked 8 Dense and MoE models \(Llama\-3, Mistral\) for BFSI compliance; implemented a Human\-in\-the\-Loop eval framework that reduced entity extraction hallucinations from 23% to 4% and improved F1 from 0\.72 to 0\.89\. • Led monthly business reviews with C\-suite to drive AI strategy, demonstrating 35% decision\-making efficiency gains from GenAI integration\.

Data Scientist at Arya\.ai \- An Aurionpro Company, Sep 2024 \- Apr 2026 · 1 yr 8 mos, Mumbai, • Scaled Bank\-Statement\-Analyzer API to 18 countries; processed 1M\+ documents/month with 98% extraction accuracy, displacing 40% of manual review load\. • Embedded on\-site with Axis Bank as end\-to\-end owner for high\-throughput document/KYC ML APIs \(pension extraction, KYC extraction, voucher/cheque classification, stamp detection, photocropping\); designed, implemented, and Dockerized each service from scratch, scaling individual APIs to ˜10,000 requests/day while maintaining latency and accuracy SLAs in Axis production infra\. • Fine\-tuned Llama\-3\.1 adapters on curated synthetic data \(distilled from larger models \+ expert validation\); optimized serving via vLLM/PagedAttention to slash P99 latency by 22% on production GPUs\.

Data Scientist at Culturelytics Pvt\. Ltd, Feb 2023 \- Aug 2024 · 1 yr 7 mos, On\-site, \* Pioneered 'People Speak' product development, specializing in sentiment analysis of employee reviews\. \* Spearheaded the collection of 2\.5Mn\+ Employee Reviews across 250\+ Employers through web scraping techniques, Which resulted in benchmarking sentiments across different sectors\. \* Implemented advanced NLP techniques and Transformer models, such as BERT and RoBERTa, to analyze and deliver sentiment trends for clients\. \* Developed interactive dashboards uncovering sentiment score and cultural factors for actionable insights\. \* Collaborated with consulting teams to deliver data\-driven recommendations to clients such as L&T, Suzlon etc\.

Data Scientist at Analytics Vidhya, Aug 2022 \- Jan 2023 · 6 mos, India, \*Implementing Predictive Modeling concepts in Python through hands\-on experience on ML projects such as customer churn prediction and loan eligibility prediction, measuring success through accuracy metrics\. \* Conducted Data Preprocessing and Exploratory Data Analysis\(EDA\) on multiple datasets, effectively managing numerous outliers and missing values to enhance data quality and analysis results\. \* Working on YouTube channel data of the company to generate useful insights and trends to build a channel growth strategy\.

Heating Ventilation Air Conditioning Intern at ISHRAE, Aug 2021 \- Oct 2021 · 3 mos, Bengaluru, Karnataka, India

Head of Braking System at Team Horus, Aug 2019 \- Mar 2022 · 2 yrs 8 mos, Bengaluru, Karnataka, India

## Education

Bachelor's degree, nitte meenakshi institute of technology

[View the original profile](https://tal.club/meet/nikhil-raj-shetty)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/nikhil-raj-shetty#person",
      "name": "Nikhil Shetty",
      "url": "https://tal.club/meet/nikhil-raj-shetty",
      "image": "https://tal.club/meet/nikhil-raj-shetty/profile-photo.jpg",
      "description": "Meet Nikhil Shetty, Production AI Systems for BFSI | Agentic RAG, Document AI, LLM Evals, vLLM/K8s | Building AI Systems Banks Can Trust based in Bengaluru. AI stack includes anthropic apis, codex, and hermes agent.",
      "jobTitle": "Senior Data Scientist",
      "worksFor": {
        "@type": "Organization",
        "name": "Arya.ai - An Aurionpro Company"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Senior Data Scientist"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Bengaluru"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "nitte meenakshi institute of technology"
        }
      ],
      "knowsAbout": [
        "anthropic apis",
        "codex",
        "hermes agent",
        "hugging face",
        "langchain",
        "langgraph"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/nikhil-raj-shetty#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/nikhil-raj-shetty#webpage",
      "url": "https://tal.club/meet/nikhil-raj-shetty",
      "name": "Nikhil Shetty on Tal",
      "description": "Meet Nikhil Shetty, Production AI Systems for BFSI | Agentic RAG, Document AI, LLM Evals, vLLM/K8s | Building AI Systems Banks Can Trust based in Bengaluru. AI stack includes anthropic apis, codex, and hermes agent.",
      "dateModified": "2026-09-03T05:48:24.445Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/nikhil-raj-shetty#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/nikhil-raj-shetty/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/nikhil-raj-shetty/profile-photo.jpg",
        "caption": "Nikhil Shetty profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```