Neelesh K Shukla Follow Scholar Medium
Principal Applied Scientist | Applied Sciences Lead – GenAI & Advanced AI @ Oracle | Former Head of AI Research & Innovation – State Street | Research Fellow | IIT Guwahati (Rank 1)
Advancing next-generation GenAI through research and enterprise-scale AI systems for reasoning, retrieval, NL2SQL, agentic memory, and Agentic AI — with a focus on scalable SLMs, feedback-driven learning, grounded reasoning, and reliable alignment.
Short Bio
Neelesh is a Principal Applied Scientist and Applied Sciences Lead at Oracle Generative AI, specializing in Generative AI, Natural Language Processing (NLP), Large Language Models (LLMs), and intelligent enterprise AI. With over 15 years of experience across scientific research, applied AI, and large-scale production systems, he leads applied science for enterprise GenAI and agentic AI capabilities, including Data Assistant/NL2SQL, Knowledge Assistant, semantic metadata systems, conversational memory, evaluation, and feedback-driven learning.
His work translates product gaps into reusable, production-grade capabilities for reasoning, retrieval, NL2SQL, agentic systems, and evaluation-driven AI, grounded in scientific rigor and real-world deployment. Before joining Oracle, he spearheaded AI research and innovation at State Street, advancing Generative AI and intelligent enterprise systems across financial services. His research spans Document Intelligence, Machine Reading Comprehension, Retrieval-Augmented Generation (RAG), Visual NLP, learning with limited labeled data, and enterprise-grade NL2SQL systems.
He contributes to the AI research community through publications and by serving as a reviewer for leading conferences and journals, including ACL Rolling Review (ARR), EMNLP, AACL, and IEEE venues. Neelesh holds an M.Tech in Computer Science and Engineering from IIT Guwahati, where he graduated as Department Rank 1, and was selected for the Government of India's prestigious international master's fellowship program associated with Heidelberg University, Germany.
Experience
Principal Applied Scientist / Applied Sciences Lead — GenAI & Advanced AI
- Lead applied science for Oracle across enterprise GenAI and agentic AI capabilities, including Data Assistant/NL2SQL, Knowledge Assistant, semantic metadata systems, conversational memory, evaluation, and feedback-driven learning.
- Define applied-research direction and convert product gaps into scientific programs, reusable architectures, benchmarks, evaluation methods, and production capabilities.
- Guide work on schema linking, ambiguity resolution, NL2SQL self-correction, reference-free answerability estimation, RAG evaluation, agent memory, and grounded reasoning over structured and unstructured data. Establish evaluation and release mechanisms covering retrieval quality, answer faithfulness, task success, trace analysis, safety, latency, cost, and production regressions.
- Partner with product organizations, engineering leaders, customers, universities, and external technology partners to validate problems and transfer research into products.
- Provide technical direction and mentorship across applied scientists and engineers while aligning research, product, architecture, and delivery priorities.
- Drive research aligned with product needs through papers, patents, benchmarks, university engagements, and external scientific collaboration.
Senior Manager, AI Group & Head — AI Innovation & Research
Incubated and led AI Research & Innovation, driving NLP and GenAI research for financial applications. Published four papers in two years. Led GenAI adoption in the Alpha Data Platform for investment management.
- Projects centered on Document Intelligence, Machine Reading Comprehension, reasoning, and question answering. Solutions included answering checklist questions from documents with evidence, financial report summarization, financial document extraction, and incident report quality checks.
- Coordinated AI Innovation and Research, focusing on financial research through publications and other engagements; evaluated state-of-the-art GenAI and LLM advances for financial use cases and State Street-specific business needs.
- Participated in the University Outreach Program, serving as a hiring manager for internships and lateral hiring, and as a member of the CTO Innovation Team.
Senior Data Scientist (SMTS)
Built an Automated ML platform, owned IoT Analytics offerings for predictive analytics and anomaly detection, and initiated research collaborations with Oracle Labs. Built end-to-end ML pipelines on Spark and big-data technologies.
- Built an Automated Machine Learning Cloud Platform to solve use cases across different domains.
- Served as the data science owner for IoT Analytics offerings, including predictive analytics, time-series analysis, anomaly detection, and user feedback.
- Started research and innovation efforts in the Oracle Analytics Product Development team and initiated collaboration with Oracle Research Labs on complex, unsolved problems.
- Partnered with product managers and customers to identify business needs, define potential solutions and KPIs, and align stakeholders before development.
- Built end-to-end data science experiments and pipelines and presented results.
- Designed end-to-end technical solutions and implemented application pipelines on Spark and other big-data ecosystem technologies.
- Self-initiated a data bank of public and in-house datasets as a common repository for use cases, promoting reuse and reducing the time needed to identify datasets and obtain legal approvals.
Visiting Research Fellow — PVS Lab
- Received an invitation to Heidelberg University, Germany, through the Department of Biotechnology (Government of India) and Heidelberg University's Visiting Master's Research Fellow program for a two-month research fellowship.
- Worked under the supervision of Prof. Dr. Artur Andrzejak, Head of the Parallel and Distributed Systems Group at the Institute of Computer Science, conducting research on code-snippet extraction.
Research Intern — Watson Data Platform
- Enhanced the IBM Watson Data Platform with cognitive-support features that enable customers to perform advanced data analytics.
- Developed methods for Shaping Function and Next Action Recommendation using data mining, language modeling, and deep learning techniques.
- Conducted comparative studies to evaluate the effectiveness of different approaches.
Senior Software Engineer — IoT Energy Management
Built an IoT-based intelligent energy-management product. Owned text search, machine-learning-based load forecasting, and the application backend. Led a team of six.
Product Developer — Banking Product
Engineered the middleware layer for the Accounts and Cards modules of the Celeriti banking product.
Systems Engineer
Led middleware development for Global Procurement, Employee Performance, and Recruitment systems. Managed a seven-member team and architected the Vendor Management System.
Highlights
Enterprise NL2SQL Orchestration
Semantic schema linking, ambiguity resolution, grounded generation, and production constraint handling for natural-language-to-SQL systems at scale. Supports Data Assistant workflows with NL2SQL self-correction and reusable schema intelligence.
RAG Evaluation Framework
Retrieval diagnostics, grounding analysis, hallucination tracing, and enterprise-grade evaluation loops for production RAG systems. Supports Knowledge Assistant capabilities through faithfulness, answerability, and trace analysis.
Agentic AI Systems
Conversation memory architectures, tool-use orchestration, multi-agent coordination, and autonomous reasoning for enterprise AI workflows. Applies conversational memory and tool orchestration to enterprise agentic AI capabilities.
Document Intelligence Pipeline
Multimodal document understanding, PDF parsing, layout analysis, and cross-lingual extraction for complex business documents. Extends to semantic metadata systems that ground reasoning over structured and unstructured data.
Small Language Model Architecture
Efficient SLM design for enterprise deployment: distillation, quantization, pruning, and task-specific fine-tuning with minimal compute footprint. Informed by feedback-driven learning and task-specific evaluation for enterprise use cases.
Learning from Feedback & Alignment
RLHF, preference optimization, reward modeling, and alignment techniques for next-generation GenAI that learns reliably from human feedback. Includes release-quality evaluation across task success, safety, latency, cost, and production regressions.
$ cat active_research_topics
$ echo $CURRENT_FOCUS
Enterprise-grade GenAI grounded in reasoning, retrieval, NL2SQL, Agentic AI, SLMs, and learning from feedback.
$ _
Publications
Patents
Awards & Honors
Education
M.Tech Computer Science & Engineering
IIT Guwahati
Department Rank 1 · CGPA 9.7/10
Thesis: Entity Role Detection in Attack Events
Research Fellowship
Heidelberg University, Germany
Government of India Master's Fellowship
PVS Lab · Prof. Artur Andrzejak
B.Tech Computer Science
MMM Engineering College