I design and develop production-focused AI automation, document intelligence, and enterprise LLM applications—from architecture and validation to deployment.
Systems designed around actual business processes and user requirements.
Independent engineering across AI, backend, interfaces, databases, and deployment.
Experience connecting modern AI applications with Oracle-backed environments.
Validation, human review, testing, and deterministic safeguards where reliability matters.
Three end-to-end projects spanning accounting automation, enterprise information access, and document-to-ERP workflows.
An OCR-powered accounting workflow that converts scanned cash vouchers into structured, reviewable records before approved data enters an Oracle-backed system.
Staff had to read each physical voucher, enter multiple fields, verify the information, and maintain an accurate history.
Scanning, OCR extraction, field detection, human validation, approval, history, and database storage in one controlled flow.
Primary solo ownership across product, OCR, backend, interface, Oracle integration, deployment, and debugging.
Time reductions are workflow-based estimates from development testing. Formal organisation-wide savings have not been measured publicly.
I work across the full product surface—from understanding the operational problem to building the AI pipeline, interface, backend, validation controls, and enterprise integrations.
Discuss your workflowOCR and extraction systems that transform vouchers, invoices, PDFs, and scanned forms into structured, validated business data.
Explore Accounts AIGoverned LLM applications that help teams retrieve business information through natural language while keeping data access controlled and testable.
Explore AJSMGPTApplications that connect repetitive operational steps, apply business rules, reduce manual handling, and introduce human approval where necessary.
Explore workflow automationComplete user-facing AI applications covering interfaces, APIs, AI components, databases, validation, deployment, and ongoing improvement.
View all projectsA controlled enterprise AI assistant that translates business questions into validated Oracle ERP queries—combining deterministic routing with evaluation, regression testing, human approval, versioned publishing, and rollback.
Generated behaviour is constrained by verified routes, read-only access, regression tests, human approval, versioned publishing, and rollback controls.
“Show the latest approved purchase for this material.”
Regression passed74 / 74 cases
Human approvedVersion 06
A controlled document-to-ERP system that identifies estimation, mixing-sheet, and purchase-order PDFs, matches related files, validates duplicates, groups records, and stores approved documents.
Handles PDFs containing multiple pages and document identifiers.
Connects estimation sheets, mixing sheets, and purchase-order files.
Checks existing records before controlled database insertion.
Integrates processed documents with the existing ERP environment.
Reduces a multi-step manual process to seconds or a few minutes per batch, depending on document size and review requirements. Formal production-wide savings are not yet measured.
Explore Estimation AII treat AI as one component of a complete product—not as a replacement for business rules, validation, or accountable decision-making.
Study the workflow, user responsibilities, failure points, data sources, and operational constraints before choosing technology.
Define where AI is useful, where deterministic logic is safer, and where human review must remain part of the process.
Develop around real scenarios, test important behaviour, expose uncertainty, and make correction straightforward.
Integrate with the existing environment, monitor usage, fix recurring failure patterns, and extend the system carefully.
Across these projects, I have worked from operational discovery and architecture through AI development, backend engineering, interface design, database integration, testing, deployment, and debugging.
I primarily built each system independently while collaborating with domain users to understand business requirements and validate correctness.
I’m Tarun, an AI Systems Engineer based in India. My work sits at the intersection of AI engineering, backend architecture, enterprise databases, and product development.
I focus on the difficult part of AI projects: understanding business rules, controlling uncertainty, integrating with existing systems, and making the final product usable by real people.
Currently a 2nd-year B.Tech student in Computer Science, specialising in Artificial Intelligence and Machine Learning—alongside building three separate AI projects for one client, supporting real manufacturing and accounting workflows.
Built through direct collaboration with domain users in a textile-manufacturing environment, using real workflow requirements, operational constraints, and testing feedback.
Client identity, production information, and identifiable business data are withheld for confidentiality.I’m available for selected AI automation projects, remote AI engineering opportunities, and technical collaborations involving practical, production-focused systems.
Whether you need to automate document processing, build a governed enterprise assistant, or connect AI with an existing business workflow, I’d be interested in understanding the problem.