Principal Data Scientist with 11 years of experience in applied LLM research and Generative AI solution design across banking, insurance, and financial services. Specialized in building science modules for RAG, NL2SQL, NL2API, and LLM evaluation, with focus on model experimentation, benchmark validation, retrieval quality, schema reasoning, and execution-accuracy assessment.
At Oracle, developed and validated domain-specific AI capabilities across 5 industry verticals, enabling MLE teams to convert research-backed modules into SDK release components.
Published applied LLM research at COLING 2025 and IEEE BigData 2025, with related patent applications filed at Oracle.
Generative AI
Agentic AI
RAG & Retrieval
Document AI
Frameworks & Tools
UI / UX
Research Interests
Oracle
Apr 2024 – Present · Bengaluru, India
Principal Data Scientist
Oracle Industries GenAI SDK Platform Mar 2025 – Present
NL2SQL Module
Designed & validated NL2SQL pipeline across 5 industry verticals
(Financial Services, Digital Government, Hospitality, Health, Construction & Engineering).
Pipeline: description enhancement → group selection → RAG+graph table selection → column selection → ambiguity detection → SQL generation → evaluation → narrative explanation.
Benchmarks: 75% Digital Govt · 78% Hospitality · 90%+ Financial Services · Spider: 72.07% (competitive with LLM baselines)
Pipeline: description enhancement → group selection → RAG+graph table selection → column selection → ambiguity detection → SQL generation → evaluation → narrative explanation.
Benchmarks: 75% Digital Govt · 78% Hospitality · 90%+ Financial Services · Spider: 72.07% (competitive with LLM baselines)
NL2SQL Studio
Interactive UI: upload catalog/schema sheet → live SQL generation & result preview,
enabling real-time pipeline validation without engineering support.
NL2API Module
API management layer (Postman → structured JSON) + working & episodic memory layers
for natural-language-driven API reasoning across Oracle Industries products.
Document AI — Form Classifier
Dual approach: (1) OCR + Multimodal LLM (2) SFT-based LayoutLM fine-tuned on GPU.
Dataset: 1,100 balanced docs — RVL-CDIP · FUNSD · CORD · PubLayNet · US/CA govt forms.
Dataset: 1,100 balanced docs — RVL-CDIP · FUNSD · CORD · PubLayNet · US/CA govt forms.
OCI GenAI Outbound — Customer GenAI Adoption Apr 2024 – Feb 2025
Major Indian Bank — Custom RAG
RAG over 4,000+ PDFs (policies, schemes, guidelines, FAQs) ·
hybrid retrieval with retriever-reranker · OCI Object Storage · top-k tuning.
Result: 85% recall on customer gold dataset before handoff.
Result: 85% recall on customer gold dataset before handoff.
Singapore Insurance Provider — Custom RAG
RAG over 16 policy/FAQ docs · evaluated on customer gold dataset ·
~90% precision on small validation set.
Oracle Financial Services — Customer Support API Suite
Talk-off Script Generator API (account JSON + loan history → structured agent guidance) ·
Customer Agent Log Report API (trends · anomalies · performance · sentiment).
Oracle Financial Services — Auto-Rules Extractor
Automated MLA/SCRA regulatory rule extraction → structured JSON for lending/leasing product integration.
Infosys
Jan 2022 – Mar 2024 · Bengaluru, India
Senior Consultant — Client: Morgan Stanley
FAST — GenAI Document Search
Semantic search across 100,000 documents for
18,000 financial advisors · ~90% efficiency ·
Featured in OpenAI customer story.
Stack: GPT-4 · MongoDB · Redis · SOLR · Kafka · CI/CD
Stack: GPT-4 · MongoDB · Redis · SOLR · Kafka · CI/CD
ChatGPT-like playground + NL-to-SQL tool (NL → SQL → NL answer) deployed firm-wide ·
GenAI POCs for BofA · American Express · Wells Fargo and others.
Accenture Solutions
Jan 2019 – Jan 2022 · Bengaluru, India
Applications Development Team Lead — Client: AT&T Communications
AI-based forecasting & anomaly detection across 2,700+ tax areas
& 50,000+ customers using time-series modelling and Flask APIs.
ML test coverage tool (sklearn · XGBoost · pandas) · REST API integration with Kronos.
ML test coverage tool (sklearn · XGBoost · pandas) · REST API integration with Kronos.
Tata Consultancy Services
Oct 2014 – Jan 2019 · Kochi, India
Systems Engineer — Client: International Monetary Fund (IMF)
PeopleSoft HR applications & enterprise reports (SQR · BI Publisher · Crystal Reports) ·
requirements analysis · performance tuning · production support.
[1]
KULFi Framework: Knowledge Utilization for Optimizing Large Language Models for Financial Causal Reasoning
COLING 2025, FinNLP Workshop
[2]
GraphRAG Analysis for Financial Narrative Summarization and A Framework for Optimizing Domain Adaptation
COLING 2025, FinNLP Workshop
[3]
Ontology-Driven Multimodal Framework for Automated Interpretation and Description of Architecture Diagrams
IEEE International Conference on Big Data (BigData) 2025
★
INSTA Award — Infosys, 2022
GenAI Document Search implementation for Morgan Stanley
★
Maestro Pinnacle Award — Accenture, 2020
Best performer at IG level
★
APEX Award — Accenture, 2020
Best performer at DU level, awarded twice
★
HR Innovation Challenge Award — Accenture, 2021
AI innovation ideas
BE, Electronics & Communication Engineering · 2014
Sri Krishna College of Engineering and Technology, Coimbatore · GPA: 7.7/10