RiturajKulshresth

ACME · Designed the four-resource credentials and authentication model (Team / App / Agent / Sub-agent) with separate edit and invocation ACLs, dual auth modes (act_as_app for production via API keys + AWS Secrets Manager, act_as_user for the playground via Okta OIDC), and a sub-agent safety property where only the caller's identity flows through every tool call - making cross-team agent reuse safe by construction.

ACME · Built the Playground for prompt-testing agents and inspecting tool calls; refactored agent configuration off local YAML onto a versioned PostgreSQL JSONB registry with a draft / published / archived lifecycle; shipped multi-model serving via AWS Bedrock alongside OpenAI and Anthropic providers.

ACME · Drove SAST findings from CRITICAL to zero in a single sprint, and bootstrapped the BOLT dual-language monorepo (FastAPI runtime + Next.js console) on Terraform / Terragrunt and Helm with CI/CD across DEV, INT, STG, and PROD.

Knowledge Hub · Designed and built a self-serve retrieval service on AWS OpenSearch where teams create and share document and index catalogs. Ingestion from Confluence, Jira, GitHub, S3, and Airflow with semantic chunking, plus multimodal embeddings via Bedrock Titan and Anthropic Claude. Authored the org-wide architecture document.

Daisy · Multi-agent platform used by 11+ teams for domain-specific chatbots. Cut monthly infra spend from ~$700 to ~$200 per non-prod environment and ~$1000 to ~$800 in PROD; built an evaluation framework with DeepEval improving evaluation accuracy by 27% and reducing chat-session storage by ~66%.

Daisy · Added visualization features for generating custom charts from Databricks and Snowflake data, and enabled ingestion of .ipynb notebooks for internal documentation.

Shipped the S2SACC Outlook add-in using Yeoman and MS Graph API, with role-based access via distribution groups, serving 1000+ users.

Built the S2EDS Outlook add-in for sending documents to enterprise servers from the mailbox, using React and Yeoman, serving 500+ users.

Migrated services from Mule ESB to AWS Lambda, reducing compute cost and improving performance for the document archival system. Maintained four critical Java applications keeping core business operations running.

Built the UI for the Document Upload interface of the DocuEdge Archiving Solution using AngularJS and Syncfusion, in production at two client organisations.

Implemented automated daily health-check procedures in the SystemWare environment using PowerShell and mainframe scripting, saving the team ~15 hours per month.

Spearheaded the development of 70+ scripts to securely migrate and restore 4000+ documents from Mobius to SystemWare ContentCloud using JCL on the mainframe.

Led the upgrade of the client's OpenText landscape and played a pivotal role in the enhancement of 10 servers.

Streamlined consumer onboarding via the Oyo Self Onboarding platform, improving the Android app user experience.

Edge-based knowledge inference for InVANETs - reduced data transferred between vehicles and Road-Side Units using NS-3 and SUMO, minimising RSU processing power needs by ~71% and enabling calculations regardless of network strength.

Built a 32-bit x86 operating system from scratch as a self-driven project: bootloader, paged memory, interrupt handling, keyboard input, a VGA output driver, and a functional shell - written in C and Assembly.

Implemented CNN models in TensorFlow that predict core body temperature from diverse facial image data (NIR, hyper-spectral, RGB), achieving 60% accuracy on a custom dataset captured under controlled lighting.

Modelled hashtag-driven social networks on Twitter and Koo using Selenium and Pandas - graph structure, cascade depth, and influencer reach for popular societal hashtags.

Built a MATLAB application for ECG signal analysis, heart-rate detection, and arrhythmia classification during a biomedical instrumentation course.