Sangeethkumar

• Lead an engineering function of 30+ engineers, owning delivery, reliability and the growth of the people in it. Led the function through a period of turnover.
• Leading the function's move to AI-native execution across architecture, development, QA, debugging and operations without mandates, and measuring where individual gains stop short of organisational ones.
• Personally built and shipped two production AI systems — the knowledge assistant and the work-progress platform — while running the function, and led a configuration-analysis service into a live customer approval path, called when a request is submitted and before it is approved.
• Designed a governance model for the context feeding an LLM: automated jobs build candidates only and never write to live, a contract gate blocks unsafe changes, humans promote, pinned versions roll back in one click. Made candidate-only a hard rule after an ungoverned nightly sync cut a production index by more than three-quarters.
• Established evaluation-first development — then found the evaluation itself was unreliable, with enough run-to-run variance that every apparent gain in an optimisation sweep sat inside the noise band. Moved decisions to offline replay over already-judged answers, confirmed with multi-run significance testing.
• Built a capacity and demand model that turned "the team feels underwater" into a month-by-month backlog projection for a leadership briefing — planned against effective capacity rather than headcount, because those are different numbers. Then scored the forecast against actuals, found it over-pessimistic, and traced why.
• Led a platform-wide enterprise-readiness assessment across dozens of services, producing an executive-ready reliability roadmap, and root-caused a severe production CPU-spike incident.
• Initiated and led a company-wide AI hackathon — ten days, real product and engineering problems, teams asked to deliver working solutions — deliberately aimed past code generation at architecture thinking, debugging and problem-solving. Set review expectations for AI-generated code: mandatory self-review, TDD-first, design review before implementation.
• Orchestrated a cloud-storage cost optimisation, delivered through the team, worth a five-figure annual saving.

• Owned the Workflow Management product end to end — data modelling, workflow automation, technical direction, and analytics delivered directly on the application platform.
• Worked on complex MongoDB queries behind the product's analytics.
• Led the team's technical decisions.

• Built the Case Management product 0→1 — product analysis, data model, system design, real-time sync, and the state machine that moves every case through its lifecycle — then led the team's technical decisions and design reviews.
• Used Google Firebase for real-time event sync in the product.
• Led the team and worked on its technically complex features.

• Built the platform's complete audit system and per-user activity feed; owned search and notification delivery across the product.
• Worked on search with Elasticsearch and notification delivery with AWS SES.

• Cost and compliance optimization and recommendation for a multi-cloud governance platform.

Python · MongoDB · AWS

• Built an automated infrastructure-approval portal for multi-cloud provisioning from scratch.
• Brought multiple cloud infrastructures into one portal, with an automated approval process for infrastructure requests from IT teams.
• Worked on Node.js as a backend developer.

Node.js · MongoDB