Entry

Score

Action

Category / Wining Level

Final Score

Validated Score

Judging Score

Pending

Information Technology - Artificial Intelligence
-

PO#

#TBA05337

Payment Status

Paid

PO Created Date

2025-12-18 02:08:11

Paid Date

2025-12-18 02:08:53

Competition

2026

Season

Entry ID

TBE105960

Entry Submission Date

2025-12-15 21:11:22

Member's Name

Sudhir Vishnubhatla

Member's Email

sudhirs.vishnubhatla@gmail.com

Entry Level

Information Technology

Results Publish

Pending

Private Entry

Yes

Hide Entry

Yes

Review

Entry Status

Category

View Entry Info

IT26008 - Artificial Intelligence

Title

Company

Entrepreneur/ Executive’s Name

Submission Type

Country

Client

Linkedin Profile Link

Project Description

Project Description for Judges

Question 1 : Provide a brief description / background of your nominated organizational, departmental, team, or individual circumstance. (Your response to this question will be used for public display if your entry is awarded.)

Sudhir Vishnubhatla is a senior technology leader and architect with over 18 years of professional experience across banking, financial technology, media analytics, and enterprise information systems. His work focuses on designing intelligent, cloud-native, and data-driven platforms that operate at scale while meeting strict requirements for security, governance, and regulatory compliance.
Sudhir currently serves as a Senior Technical Lead at Infinite Computer Solutions, where he architects enterprise-scale AI and cloud platforms for regulated environments. He designed a multi-agent decision ecosystem for Intelligent Document Processing (IDP), integrating OCR, NLP, and transformer-based models to enable automated classification, extraction, calibration, routing, and governance. His architecture emphasizes responsible AI, with built-in auditability, explainability, and confidence-driven decisioning to support compliance-critical workflows.
Previously, Sudhir worked at Nielsen Media Research, where he led large-scale cloud migration and big data modernization initiatives, transitioning high-volume on-premises analytics platforms to AWS. He architected a fully open-source Business Intelligence platform to replace commercial BI tools, significantly reducing licensing costs while improving scalability and maintainability. He also designed custom Apache Spark clusters and introduced AI-driven optimization techniques to improve performance, reliability, and cost efficiency across global data pipelines.
Earlier in his career, Sudhir held engineering roles at Randstad, Cognizant, and IBM, where he contributed to the development and modernization of enterprise systems for financial services, analytics, and document generation. These roles provided him with a strong foundation in Java, distributed systems, big data processing, and enterprise software engineering.
In addition to industry leadership, Sudhir is an active contributor to the research community. He has published multiple peer-reviewed journal articles and a book chapter on applied AI, intelligent information management systems, big data governance, and ethical AI, and he regularly participates in journal peer review and editorial activities. He is also a co-inventor of a UK design patent focused on AI-powered process optimization.
Sudhir holds a Bachelor’s degree in Computer Science & Engineering and maintains industry-recognized certifications, including Google Certified Professional – Cloud Architect, NVIDIA Certified Professional – GenAI LLMs, and Cloudera Certified Spark & Hadoop Developer. His areas of expertise include Artificial Intelligence, cloud migrations, Big Data, data engineering, and enterprise platform modernization.

Question 2 : Outline your nominated organizational, departmental, team, or individual achievements, dated 1-3 years prior, based on your selected categories.

Over the past two years, I designed and delivered a compliance-grade, multi-agent Artificial Intelligence platform that enables large-scale automation in regulated enterprise environments without losing transparency, control, or trust. This work addressed a key challenge in enterprise AI: scaling automation while preserving explainability, auditability, and predictable operating cost.
The platform currently supports high-volume Intelligent Document Processing workflows across financial services, tax processing, and auto finance, with a clear path for future expansion through an Extensible Agent Foundry (EAF). It processes more than 125,000 documents per day with stable performance and controlled cost. By engineering AI decision-making as a governed system rather than a single model, the platform improved Straight-Through Processing (STP) from 18% to over 62%, and cut manual review effort by 50–80%.
Through small-model-first optimization and cost-aware routing, the system achieved 40–60% reduction in AI compute usage while maintaining day-one accuracy above 90% for new document types and sustained accuracy for mature partners. These optimizations enabled faster turnaround times and same-day decisions for high-volume workflows.
To future-proof the solution, I also designed the Extensible Agent Foundry and Cognitive SDK. This framework abstracts the event-driven orchestration layer and shared knowledge graph, allowing teams to rapidly build and plug in new domain-specific agents without impacting system stability. As a result, the platform can be extended into new domains like compliance research, or regulatory monitoring, without re-architecting the core system.
Beyond architecture and engineering, I led end-to-end delivery of the initiative, owning technical planning, milestone definition, and cross-team coordination. I worked closely with compliance, business, and operations stakeholders to translate regulatory and operational requirements into production-ready designs, ensuring alignment across teams and timely execution. This leadership reduced rework and helped in faster adoption of responsible AI practices.
I published multiple peer-reviewed and applied research articles on agentic AI, ethical document automation, multilingual LLMs, and AI governance, directly reflecting the production architecture. I also earned Google Cloud Professional – Cloud Architect and NVIDIA Certified Professional – GenAI LLMs certifications to ensure my designs align with industry best practices.

Question 3 : Provide examples of the strategies / plans you have introduced, in order to attain your nominated organizational, departmental, team, or individual achievements. Explain why the highlighted achievements are significant / unique.

To achieve these results, I designed a multi-agent AI system that breaks complex workflows into clear, accountable steps, validates every decision, and enforces regulatory rules. Instead of relying on a single, monolithic AI pipeline, the platform uses specialized agents coordinated through an asynchronous, event-driven orchestration layer, enabling parallel processing, and controlled routing.
A core strategy was diving the system into clearly defined agents, ensuring each performs a single, well-defined function to minimize error propagation. The Classification Agent identifies document subtypes using a combination of text cues, visual layout analysis, and structural patterns, achieving over 99% accuracy. This prevents downstream errors and avoiding unnecessary AI calls. The Extraction Agent uses quantized Small Language Models (SLMs) and multimodal layout parsers to extract structured fields, tables, and signatures. By applying document-focused retrieval and structured output enforcement, input token usage was reduced by nearly 70%, significantly lowering cost and improving latency.
Trust and compliance are maintained through the Calibration Agent, which generates confidence scores using five validation layers: OCR quality, deterministic schema checks via Postgres, semantic alignment using Neo4j, contextual completeness, and internal consistency. Every extracted value is scored and auditable. The Governance Agent combines these confidence signals with business rules, ontology checks, and anomaly detection to decide whether a document qualifies for straight-through processing or requires human review. This design improved STP from 18% to 62% and reduced exception queues by 40% in high-volume financial workflows.
The Evaluation Agent applies IRS tax rules, loan conditions, and industry-specific validations using rule engines and graph lookups, ensuring decisions are not only accurate but compliant. Continuous improvement is enabled through the Feedback Agent, which incorporates human corrections into templates, entity mappings, and SLM prompts without costly retraining.
To ensure the platform remains adaptable over time, I introduced an Extensible Agent Foundry and Cognitive SDK, allowing new domain-specific agents to be rapidly plugged in without destabilizing the system. Together, these strategies move enterprise AI from experimental use into production-ready, compliant systems, showing that scalable AI success depends more on system design and governance than on larger models or ad-hoc tuning.

Optional Context for Judges (Not for publication. This section is solely for judges' evaluation.)

This submission reflects practical, production-ready Artificial Intelligence built for regulated enterprise environments, where compliance, auditability, and cost control are as important as accuracy. Rather than focusing on model novelty, the work emphasizes building AI systems that can be trusted, operated, and scaled in real business settings.
This work is shaped by over 18 years of hands-on industry experience spanning Artificial Intelligence, cloud migrations, and large-scale data engineering. It shows consistent delivery of real enterprise systems
The architecture is supported by peer-reviewed and applied research publications in agentic AI, ethical document automation, multilingual LLMs, and AI governance, ensuring strong alignment between research and production implementation. In addition, advanced certifications like Google Cloud Professional - Cloud Architect and NVIDIA Certified Professional - GenAI LLMs, reflect current best practices in cloud and AI engineering and support the technical choices I have made.
I am also a co-inventor of a UK design patent related to AI-driven process optimization, which aligns directly with the intelligent, adaptive systems demonstrated here.
Overall, my work represents a proven approach for responsible enterprise AI, showing how disciplined engineering, governance, and cost control translate AI capabilities into consistent, measurable business results.

Thumbnail

Cancel

Link List

Link	Password Protected	Display (Winner Info)
https://www.credly.com/badges/f368d7cc-c4b3-4b78-9487-140a5de591e6	No	Display
https://www.credly.com/badges/5e33eafe-beeb-4b28-866a-059926972ec3	No	Display
https://globalrecognitionawards.org/winners/2025/sudhir-vishnubhatla-recognized-with-a-2025-global-recognition-award/	No	Display
https://threws.com/sudhir-vishnubhatla-pioneering-the-future-of-applied-ai-and-data-intelligence/	No	Display
https://techbullion.com/the-accountable-algorithm-how-streaming-intelligence-is-redefining-trust-in-financial-ai/	No	Display

Credit List

Actions	Name	Title Organization

File List

File Link	File Size	File Link (System Resize)	File Size (System Resize)	Display(Winner Page)
upload/entry/files/TBE105960/30781766023245.pdf	471.81KB	upload/entry/files/TBE105960/30781766023245.pdf	471.81KB	Hide
upload/entry/files/TBE105960/17971766023288.pdf	528.99KB	upload/entry/files/TBE105960/17971766023288.pdf	528.99KB	Hide
upload/entry/files/TBE105960/47601766023312.pdf	455.66KB	upload/entry/files/TBE105960/47601766023312.pdf	455.66KB	Hide
upload/entry/files/TBE105960/39341766023341.pdf	1.58MB	upload/entry/files/TBE105960/39341766023341.pdf	1.58MB	Hide
upload/entry/files/TBE105960/1769062510file_11489.jpg	228.06KB	upload/entry/files/TBE105960/1769062510file_11489.jpg	228.06KB	Hide

Activity Logs

Action	Details	Performed By	IP Address	Date
Upload	Upload single entry file	Admin: Izzah	15.204.130.138	2026-01-22 06:15:10
Edit	Update entry status from New to Approved	Admin:	15.204.166.127	2026-01-22 07:36:54
Upload	Upload entry thumbnail	Admin:	15.204.166.127	2026-01-22 07:36:54

Edit Entry

Score

Link List

Credit List

File List

Activity Logs