Akash S

Role Scope: I co-own and operate a high-volume server-side print microservice and drive safe, fast releases with an AI-enabled test framework that dark-launches n+1 alongside n. Release Quality (AI): The dark-launch validator doubled release confidence and reduced the test cycle from 30 days to 1 day (–97%). Operations & Incident Response: Golden dashboards and incident playbooks improved incident-handling efficiency by 40%. Customer Success: Coordinated upgrade programs reduced support cases by 25% post-release. Reliability & Scale: I consistently meet strict latency and availability SLOs at multi-million-requests/day scale. Platform & Cloud Modernization: I led the migration from ECS to EKS and into a Platform-Centric Accounts (PCA) model on AWS, strengthening standardization, security guardrails, and long-term scalability. Developer Community & Enablement: I actively contribute to the Windsurf community through the Windsurf Champions Program, sharing patterns and playbooks that help teams adopt best practices faster.

Role Scope: I co-owned the centralized SSO/Authorization platform and contributed to core platform services. Integrating SKUs with SSO: I standardized SKU onboarding to a Keycloak-based SSO, automating RBAC and client provisioning, which reduced time to onboard new SKUs by 50%. IdP Exploration & Pilots: I explored Keycloak capabilities and ran POCs for enterprise IdPs—PingFederate, Microsoft Entra ID (AD), OneLogin, Okta—across SAML and OAuth/OIDC, which proved decisive in retaining customers by meeting their identity requirements. RAG-based Incident Retrieval for Jira: I built a Jira-connected RAG assistant with LlamaIndex, OpenBioLLM, PGVector, and Gradio that retrieves similar incidents, surfaces code/KB/runbook snippets, and drafts fix steps—accelerating L1/L2 triage and standardizing responses. Audit & Task Management: I contributed to audit and task-management services adopted across SKUs, providing traceability and QMO/compliance readiness for Life Sciences workflows. Modernization: I led a spike and upgrade path for the SSO codebase from Java 11 to Java 17, validating feature retention and backward compatibility.

Role Scope: I built custom data integrations (Pandas/Airflow) and templatized SDQ (Smart Data Quality) checks, then helped found the Platform Engineering function. EDC Outbound (Configurable): I delivered configurable outbound connectors that reduced customer onboarding time by 20% and made deployments repeatable across accounts. Ops UI & Multi-Tenancy: I added multi-tenant support and an internal writeback status/retry UI that enabled L1 agents, cutting L2 production tickets by 75%. Demo/POC Enablement: I created sanity-check suites that kept demo environments green, letting Sales/SEs pitch 25% more customers in the same time window.

Role Scope: I implemented Workday PECI integrations for payroll vendors including PwC, ADP, and NGA across more than 80 countries, coordinating with HRIT and vendor teams through testing to go-live. Integration Delivery: I designed and delivered country-specific PECI data mappings that aligned Workday worker data to each vendor’s specification, then guided end-to-end validation with payroll stakeholders. Automation for Specs and Readiness: I built Python automation to extract representative datasets from existing systems, profile fields, and validate against vendor specification documents, which accelerated integration readiness and reduced manual rework. Quality and Consistency: I templated checklists and repeatable validation steps so new country rollouts followed the same process and achieved consistent outcomes. Collaboration and Enablement: I documented SOPs and handover notes that helped project teams onboard new geographies quickly and maintain integrations after launch.

Role Scope: I built analytics and backend components that supported COVID-19 safety operations and RFID-based telemetry in a manufacturing setting. Contact Tracing Analytics: I developed a Streamlit dashboard that visualized proximity and exposure windows for employees, enabling rapid identification of likely contacts and faster response by the safety team. RFID IoT Backend: I contributed to a Node.js REST API that ingested events from vehicle-mounted RFID devices and persisted them in SQL Server with input validation and reliable storage patterns. Cloud Deployment: I packaged and deployed applications using Azure App Service, enabling repeatable releases and straightforward rollbacks during frequent updates. Data Integrity and Access: I added basic data checks and role-based views to improve accuracy and limit access to sensitive information.

Role Scope: I built line-side utilities and prototype apps with Python and .NET in collaboration with operators. Computer Vision QA: I implemented OpenCV template matching on a live camera feed to flag misalignment and defects, tuned ROIs, thresholds, and debouncing to reduce noise, and packaged a small utility for the line. Industry 4.0 AR: I developed an augmented reality prototype for warehouse operations that improved pick and location guidance and supported Industry 4.0 readiness. Label Scanning to SAP: I implemented automated camera-based label scanners using OCR and OpenCV and integrated them with SAP data entry through a Node.js service, which reduced manual typing and sped up line operations. Docs and Enablement: I documented setup steps, calibration guidance, and failure modes so non-ML users could maintain the tools.

Dissertation: Benchmarking domain-focused LLMs for automated medical coding from clinical notes (ICD and CPT). Evaluation: Built a harness for precision@K, coverage, reviewer accept rate, and an error taxonomy covering synonym drift and laterality. Outcome: Practical guidelines on prompt patterns, retrieval window sizing, and reviewer-in-the-loop workflows for safer clinical NLP. Keywords: LLM evaluation, prompt design, small-context retrieval, healthcare NLP.

Selected coursework: Data Structures and Algorithms, Operating Systems, DBMS, Computer Networks, Distributed Systems, Machine Learning, Software Engineering.