---
title: "Rishabh Vatsa - Artificial Intelligence Engineer in Bengaluru | Tal"
description: "Meet Rishabh Vatsa, ​Building S45 - AI Native IB | AI Engineer | Applied ML & Agentic Frameworks | Ex DS @SIGN3 based in Bengaluru. AI stack includes anthropic apis, claude code, and hugging face. Explore public work links including github and medium."
image: "https://tal.club/meet/rishabh-vatsa/share-image"
---

# Rishabh Vatsa on Tal

## Headline

​Building S45 \- AI Native IB \| AI Engineer \| Applied ML & Agentic Frameworks \| Ex DS @SIGN3

## About

Data Scientist turned AI Engineer, working mostly in Indian fintech \- fraud, credit risk, and more recently AI over capital\-markets data\. Most of the work lives in the unglamorous middle: getting models and LLM systems out of notebooks and into production, where they have to stay reliable, explainable, and cheap enough to actually run\. That part tends to be harder than the modelling \- and more interesting\. At S45, the focus is the retrieval and extraction layer behind a SEBI\-regulated IPO\-intelligence platform \- getting language models to answer questions over dense regulatory filings accurately, affordably, and with sources that can be checked\. Governance \- PII handling, audit\-ability, a human in the loop \- isn't a feature here; it's the constraint everything is designed around\. Before that, a few years at Sign3 on the ML behind fraud and credit products: embeddings, transformer fine\-tuning, and a device\-fingerprinting framework for entity resolution \- plus plenty of time on the parts that rarely get talked about, like feature pipelines, evaluation, and internal tooling\. Comfortable across both classical ML and modern LLM/agentic systems, but more interested in whether something holds up in the real world than in whichever technique happens to be fashionable\. Always happy to talk shop \- retrieval, fraud and risk modelling, or what it actually takes to ship AI\.

## Location

Bengaluru

## AI stack

anthropic apis

claude code

hugging face

langchain

OpenRouter

PydanticAI

## Work experience

Artificial Intelligence Engineer at S45, May 2026 \- Present · 6 mos, Bengaluru, Karnataka, India, RAG Platform: Built the multi\-tenant retrieval system behind a production LLM assistant over Indian IPO filings, plus its LLM\-as\-Judge eval harness: 97\.5% citation\-grounded accuracy at \~25x lower cost than the LLM\-only system it replaced\. Agentic IPO Research Assistant: Production chatbot on LangGraph with human\-in\-the\-loop clarification, context\-management middleware, end\-to\-end tracing, and automatic Indian PII redaction \(PAN, Aadhaar, mobile, email\)\. Conversational Document Agent: Agentic chat app reasoning over user\-chosen documents, answering with grounded, page\-level citations\. Chart Intelligence \(Vision\): Pipeline that turns charts in industry decks into retrievable tables; 43 decks \(\~1,800 pages\) live in production; 87% on ChartWalker\-Bench vs \~64% for published multimodal systems\. Field Extraction: LLM pipeline extracting 100\+ structured fields per filing; worst\-case completeness 27% to 61%; 345\+ zero\-cost tests; provenance ledger plus guardrails that caught a silent net\-worth mis\-extraction in production\. Human Edit Control: Designed the system keeping banker edits alive in AI\-generated due\-diligence reports: edits survive regeneration, reasons upgrade into propagating rules; proved by blind multi\-agent simulation\. Due\-Diligence Reports: Deterministic tables paired with an analyst\-and\-verifier LLM loop; 200\+ faithfulness\-verified sub\-industry reports used by bankers on live deals\. LLM Observability \(OSS\): Wrote the company's AI telemetry standard and open\-sourced s45\-telemetry \(Apache\-2\.0\): agent spans to Logfire, system spans to Grafana, PII scrubbing; telemetry coverage 54% to 96%\. Document Ingestion: Fault\-tolerant parse/embed/index service on AWS ECS with retries and a dead\-letter queue; in production behind multiple systems\. Backend Re\-architecture: Single CI\-enforced data\-access boundary \(per\-aggregate interfaces, immutable DTOs\); migrated 100% of legacy SQL with equivalence tests, zero behavior change across a 200\+ file service\.

Data Scientist at SIGN3, Dec 2022 \- May 2026 · 3 yrs 6 mos, Gurugram, Haryana, India, Financial SMS Intelligence: Built a 6\-layer SMS parsing, classification and event\-resolution engine from scratch \(90\+ bank parsers, a 369\-template engine, 0\.96 weighted F1 at 50K msg/s over FastAPI\) that turns raw, duplicate\-heavy bank SMS into deduplicated, canonical financial events\. Credit\-Risk Feature Engineering: On those events, engineered base and derived behavioral underwriting features for a microfinance lender's 15\-30 day high\-risk emergency loans, such as 90\-day EMI bounces and credit\-card utilisation, the strongest reaching Information Values \(IV\) of 0\.36 and 0\.31\. MLOps & Model Deployment: Owned MLOps in the final 1\.5 years: built a modular Embeddings\-as\-a\-Service \(EaaS\) API \(FastAPI \+ PyTorch\) with model versioning and lazy loading, serving Attention Encoder models to downstream fintech scoring systems\. Finetuning & Model Training: Trained transformer models \(MLM, LoRA, supervised autoencoders\) for affluence prediction, DPD/NPA scoring, and phone\-disposability detection, achieving up to a 24% lift in AUC\. Device Fingerprinting & Entity Resolution: Productionized a quantitative device\-fingerprinting framework for entity resolution, classifying identifiers by stability and cutting the device collision rate by 81% while sustaining all known splitting cases\. Applied ML Research & Behavioral Biometrics: Built behavioral\-biometrics frameworks from scratch from ML research, applying Few\-Shot Learning and Federated Multi\-Modal Contrastive Learning to continuous authentication\. Fraud, Risk & Decisioning: Led client PoCs for fraud\-ring detection, first\-party fraud, and user segmentation across first/third\-party, digital and device data; built reusable SQL and feature stores plus grid\-search training and metric tooling \(WoE\-IV, KS, AUC\), and migrated analytics from Athena to BigQuery at full parity\.

Associate Product Analyst at TEG Analytics, Jul 2022 \- Dec 2022 · 6 mos, Bengaluru, Karnataka, India, Automated the generation of weekly RFM \(Recency, Frequency, Monetary\) reports, enhancing efficiency and accuracy in customer behavior analysis Collaborated on consolidating UI/UX enhancements for dashboards, resulting in improved user experience and engagement Designed and implemented the ETL process for monthly and annual data refresh of the dashboard, ensuring up\-to\-date and reliable insights

Data Science Intern at SIGN3, Apr 2022 \- Jul 2022 · 4 mos, Gurugram, Haryana, India, Conducted in\-depth analysis of research papers focused on keystroke and touch dynamics implicit fingerprinting techniques, contributing to a comprehensive understanding of the field Developed and refined LSTM \(Long Short\-Term Memory\) models tailored to touch dynamics implicit fingerprinting for diverse users, creating personalized personas for enhanced user identification Successfully implemented the MFk\-M clustering algorithm based on its foundational paper, specifically tailored for high\-dimensional categorical datasets

Product and Community Analyst at SIGN3, Mar 2022 \- Apr 2022 · 1 mo, Hybrid, Investigated the importance of developer\-based communities, understanding their crucial role in fostering collaboration, innovation, and product advancement of API and SDK based products Explored the initiatives and goals of the Fido Alliance, recognizing its significance in promoting secure authentication standards and contributing insights into the broader technology landscape

## Education

Bachelor of Technology \- BTech, Dr B R Ambedkar National Institute of Technology, Jalandhar

St\. Xavier's Junior/Senior School, Muzaffarpur

## Public links

- github: https://github\.com/vatsarishabh22
- medium: https://medium\.com/@vatsarishabh22

[View the original profile](https://tal.club/meet/rishabh-vatsa)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/rishabh-vatsa#person",
      "name": "Rishabh Vatsa",
      "url": "https://tal.club/meet/rishabh-vatsa",
      "image": "https://tal.club/meet/rishabh-vatsa/profile-photo.jpg",
      "description": "Meet Rishabh Vatsa, ​Building S45 - AI Native IB | AI Engineer | Applied ML & Agentic Frameworks | Ex DS @SIGN3 based in Bengaluru. AI stack includes anthropic apis, claude code, and hugging face. Explore public work links including github and medium.",
      "jobTitle": "Artificial Intelligence Engineer",
      "worksFor": {
        "@type": "Organization",
        "name": "S45"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Artificial Intelligence Engineer"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Bengaluru"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Dr B R Ambedkar National Institute of Technology, Jalandhar"
        },
        {
          "@type": "EducationalOrganization",
          "name": "St. Xavier's Junior/Senior School, Muzaffarpur"
        }
      ],
      "knowsAbout": [
        "anthropic apis",
        "claude code",
        "hugging face",
        "langchain",
        "OpenRouter",
        "PydanticAI"
      ],
      "sameAs": [
        "https://github.com/vatsarishabh22",
        "https://medium.com/@vatsarishabh22"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/rishabh-vatsa#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/rishabh-vatsa#webpage",
      "url": "https://tal.club/meet/rishabh-vatsa",
      "name": "Rishabh Vatsa on Tal",
      "description": "Meet Rishabh Vatsa, ​Building S45 - AI Native IB | AI Engineer | Applied ML & Agentic Frameworks | Ex DS @SIGN3 based in Bengaluru. AI stack includes anthropic apis, claude code, and hugging face. Explore public work links including github and medium.",
      "dateModified": "2026-10-16T08:31:37.960Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/rishabh-vatsa#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/rishabh-vatsa/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/rishabh-vatsa/profile-photo.jpg",
        "caption": "Rishabh Vatsa profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```