Apoorv Mittal

• 20+ engineers directly managed across two Lead roles. Mentored two engineers to senior-track promotions in the last 18 months at AutoScout24.
• Led a strangler-fig replatforming of the public listing-search GraphQL API at AutoScout24 onto graphql-yoga + hono, built in parallel for four months, every production query shadow-diffed against the legacy service before any traffic shift.
• Drove a 30%+ p50 latency reduction on a public API serving 50+ downstream integrations.
• Built and rolled out a suite of Claude Code agents (build, scan, PR review, migration, performance) across multiple production codebases.
• Shipped LLM features on the Anthropic API with SQL validation, parameterised queries, and RBAC scoping.
• Claude Code, Cursor, Copilot and CodeRabbit adopted across squads, with a shared Claude Code skill pack (7 skills, 5 custom commands) built from recurring review comments.
• Structured hiring: designed and calibrated hiring loops across squads; full-loop hiring with CTO/VP sign-off while scaling the CheMondis team.
• Regulated domains: insurance (Bupa International, MetLife) and FCA-regulated retail banking (Lloyds Banking Group).

People management

1:1s, quarterly reviews, IDPs, promotion cases and PIPs for 8 direct reports; input into compensation decisions; 20+ engineers managed across two Lead roles.

Structured hiring

Hiring loops with defined stages (screening, technical, system design) and clear evaluation criteria; bar calibration across squads; onboarding guides so new hires ramp quickly.

Coaching technical leads

Coach senior engineers toward staff and lead tracks. Technical decisions stay with the engineers closest to the code and are recorded in ADRs; my job is to unblock, not override.

Capacity & delivery

Quarterly OKR cycles end to end; balance roadmap demand against tech debt and reliability work; planning and ceremonies with Product and delivery partners.

Cross-boundary coordination

Lead change across squads that do not report to me: shared standards, sequencing and early dependency resolution; partner with Product, Design, Backend and platform teams and external vendors (Stripe, Trustshare, Looker, LaunchDarkly).

Governance, change & reliability

Shipped inside FCA release and compliance controls at Lloyds; own on-call, runbooks, incident response and blameless postmortems; codified guardrails and review norms for AI tools.

1. Fixed deadline, 20+ repositories, squads I didn't manage

   Situation

   The organisation's Apollo/GraphQL stack had fallen behind. Upgrading it touched 20+ repositories owned by several squads, with a fixed quarterly deadline and customer-facing traffic that could not go down.

   Action

   Split the work into staged waves, agreed shared standards and one migration tracker with each squad lead, used agents for bulk scaffolding and progress tracking, and left implementation decisions with the engineers who owned each codebase.

   Result

   Delivered on the fixed quarterly timeline with zero downtime.

2. Agentic AI as part of the engineering lifecycle, not a side tool

   Situation

   AI tools were used ad hoc across squads on our React, Next.js, TypeScript, Node.js and GraphQL codebases, with no shared norms, while the same review feedback and quality checks were repeated by hand.

   Action

   Packaged our standards as a shared Claude Code skill pack (7 skills, 5 custom commands) for PR review, pre-push gates, CI diagnostics and migration tracking, built from recurring review comments; added CodeRabbit on every PR and a tsc → tests → lint → build loop; rolled it out squad by squad with norms, guardrails and hands-on onboarding.

   Result

   Agents became part of everyday delivery across squads, from coding to review and CI; repetitive checks moved to automation, freeing engineers for design, risk and product questions.

3. Feature pressure vs. quality in a fast-growing team

   Situation

   As CheMondis scaled, feature demand kept crowding out tech debt while the engineering team was growing quickly.

   Action

   Negotiated protected tech-debt capacity with Product, Design and Backend leads; introduced SLOs with error budgets, ADRs, runbooks and onboarding guides; ran full-loop hiring with CTO/VP sign-off.

   Result

   The team grew with a shared, data-backed way for Product and Engineering to decide when to slow down.

Agent design & verification

Agents built with the Claude Agent SDK and Claude Code (custom agents, hooks, MCP servers); a tsc → tests → lint → build verification loop and a domain-scope refusal policy for agents; agents for migrations, CI-failure diagnostics and flaky tests.

LLM features in production

Anthropic API: natural-language SQL over multi-tenant PostgreSQL (SQL validation, parameterised queries, RBAC scoping), anomaly detection, AI tutor, AI-guided onboarding.

Retrieval & prompt engineering

RAG and agent orchestration with LangChain, local inference with Ollama, scoped-knowledge system prompts with prompt-injection defences.

Latency & cost control

Live assistant on motivatorpal.com: prompt caching, streaming, per-IP rate limiting, a daily cost cap and per-request token-cost tracking.

Claude on AWS Bedrock

Claude models through Amazon Bedrock: IAM-scoped access and AWS-native deployment that fits existing AWS, Terraform and Kubernetes infrastructure.

LLM observability & cost

Langfuse tracing of prompts, tool calls and agent runs with latency and token cost per feature.

Languages & APIs

TypeScript/Node.js, Python, Java (Lloyds); REST and GraphQL API design, versioning and BFF patterns; typed schemas and runtime validation.

Cloud & containers

Azure (Lloyds) and AWS (AutoScout24); infrastructure as code with Terraform; Docker and Kubernetes deployments (manifests, Helm, kubectl).

CI/CD & quality

GitHub Actions, CircleCI, Jenkins; unit, integration and E2E testing; quality gates on every PR; feature flags and staged rollouts.

Data

PostgreSQL schema design and migrations with DDL-safety checks, multi-tenant RBAC, idempotency, audit trails on money mutations; MongoDB, DynamoDB, Cassandra.

Observability & operations

DataDog and Sentry dashboards, alerting and tracing, SLAs/SLOs with error budgets, runbooks, root-cause analysis and preventative follow-ups.

Security in regulated settings

Secrets handling, least-privilege data scoping, CSP and security headers; sensitive-data handling for document capture and face match at Lloyds; traceable changes.

People leadership

8 direct reports; 1:1s, IDPs, performance reviews, PIPs; full-loop hiring; coaching to staff/lead; hybrid teams across 50+ nationalities.

Architecture & quality

Problem solving, system design, architecture reviews; rigorous unit/integration/E2E testing; WCAG 2.1 AA; DORA-oriented metrics; incident response and blameless postmortems; on-call and runbooks; CI/CD; SLAs/SLOs.

Frontend, full-stack, mobile

React 18, Next.js, TypeScript, Node.js, NestJS, Python, Java, GraphQL (graphql-yoga, Apollo), REST, Zustand/Redux, micro frontends, Storybook, Core Web Vitals, i18n/l10n, SEO, LaunchDarkly, Jest/RTL/Playwright/Cypress, Sitecore, AEM, Capacitor.

Cloud, infra, data

AWS CloudFront/S3, Azure, Vercel, Docker, Kubernetes, Terraform; Git and CI/CD (GitHub Actions, CircleCI, Jenkins); observability and production debugging with DataDog and Sentry; Stripe, Trustshare; production PostgreSQL with multi-tenant RBAC, idempotency, and SQL-safety practices; NoSQL (MongoDB, DynamoDB, Cassandra); networking and security fundamentals; Looker, Metabase.

AI / generative AI

Claude Code agents in production (build, scan, PR review, migration, RBAC validation, audit-trail wiring). LLM features on the Anthropic API with SQL validation, parameterised queries, RBAC. LangChain, Ollama for local inference. Shared Claude Code skill pack and CodeRabbit rolled out across squads. Custom agents on the Claude Agent SDK (hooks, MCP servers), Claude on AWS Bedrock, Langfuse tracing.

• People management
• Hiring & calibration
• Performance & career frameworks
• Coaching tech leads
• Capacity planning
• Stakeholder management
• Effective communication
• AI governance & guardrails
• LLM agents & RAG
• Prompt engineering
• Claude Code · Cursor · CodeRabbit · Copilot
• Claude Agent SDK · MCP
• Claude on AWS Bedrock
• Anthropic & OpenAI APIs
• Langfuse
• TypeScript
• Node.js
• Python
• Java
• React
• Next.js
• GraphQL
• REST API design
• Micro frontends
• PostgreSQL
• Docker & Kubernetes
• Terraform
• Azure · AWS
• GitHub Actions
• DataDog · Sentry
• Jest · RTL · Playwright
• A/B testing & feature toggles
• Accessibility (WCAG 2.1 AA)
• Performance
• Looker · Metabase
• Figma · Storybook

1. Aug 2022 to Present · Munich, Germany

   Engineering Lead, Full-Stack & AI Technologies

   AutoScout24 GmbH, Europe's largest online car marketplace.

   React 18 · Next.js · TypeScript · GraphQL (graphql-yoga, Apollo + Federation) · Node · Zustand · LaunchDarkly · Playwright · Python · AWS · Terraform · Docker · Kubernetes · Claude Code · Cursor · CodeRabbit

   • Engineering Lead with 8 direct reports, combining people management with hands-on architecture, code review and delivery ownership; owns 1:1s, quarterly reviews, IDPs, promotion cases, and PIPs; calibrates hiring across squads and contributes to compensation decisions.
   • Mentored 2 engineers to senior-track promotions in the last 18 months.
   • Led a strangler-fig replatforming of the public listing-search GraphQL API onto graphql-yoga + hono: parallel build for four months, shadow-diff pipeline replaying every production query against v1, two-rail schema-drift detection, custom yoga plugins for observability.
   • Orchestrated performance work on the most-consumed public API; delivered a 30%+ p50 latency reduction.
   • Established analytics-driven experimentation with LaunchDarkly; contributed to a low-single-digit retention uplift on tracked journeys.
   • Built and rolled out a shared Claude Code skill pack (7 skills, 5 custom commands) for PR review, quality gates, CI diagnostics and migration tracking, plus CodeRabbit on every PR; drove team-wide adoption of Claude Code, Cursor and Copilot with review norms and onboarding.
   • Led the organisation-wide GraphQL modernisation across 20+ repositories (Apollo upgrades, GraphQL Tools, type generation, TypeScript bumps) with a staged rollout and zero downtime on a fixed quarterly timeline.
   • Hands-on with the delivery platform: services on AWS, infrastructure as code with Terraform, containerised deployments to Kubernetes and CI/CD quality gates.

2. Oct 2019 to Jul 2022 · Cologne, Germany

   Tech Lead

   CheMondis GmbH, Europe's leading B2B marketplace for chemicals: e-commerce, payments, seller onboarding.

   React · TypeScript · Redux · Node · Styled Components · Storybook · Jest · Cypress · Sentry · DataDog · Stripe · Trustshare

   • Tech Lead with up to 5 direct reports: 1:1s, performance reviews, IDPs, onboarding and escalations.
   • Defined and monitored SLAs/SLOs for customer-facing flows and owned error-budget conversations with Product; production monitoring via DataDog and Sentry.
   • Full-loop hiring with CTO/VP sign-off; scaled the frontend team as the business grew.
   • Architected the foundational frontend platform: component library, list/filter/search modules, analytics infrastructure.
   • Shipped commerce capabilities: Stripe payments, Trustshare escrow, GDPR cookie consent, Looker analytics, MJML transactional email.

3. Jan 2016 to Mar 2019 · London, UK

   Senior Interactive Developer

   Publicis Sapient (client: Lloyds Banking Group), FCA-regulated retail banking with 30M+ customers.

   Java · JavaScript/TypeScript (ES6+) · React · Redux · Azure · Docker · SCSS · Jest · Jasmine/Karma · Sitecore · SCEP component library

   • Delivered a reusable document-capture and face-match module and Financial Wellbeing platform features for retail banking onboarding, within FCA-regulated security and release constraints.
   • Worked beyond the UI on Java services and Azure-hosted environments supporting the onboarding journeys, inside the bank's change and release controls.
   • Core contributor on SCEP, the unified component library powering Lloyds' public and secured web estate.
   • Established WCAG 2.1 AA and legacy-browser (IE9+) support standards. Active in hiring loops.

4. 2012 to 2015 · Gurgaon, India

   Interactive Developer (Publicis Sapient, 2014–2015) · System Engineer (TCS, 2012–2014)

   Publicis Sapient · Tata Consultancy Services, Bupa International, MetLife China, Fidelity Worldwide Investments, HP.

   Sitecore · Salesforce · Cross-platform mobile

   • Sitecore component library and bilingual web platforms with integrated payments.
   • Cross-platform mobile delivery and a Salesforce-based B2B gateway.

English

Full professional

Hindi

Native

German

A2, improving

Science fiction & tech trends · Travelling · Table tennis & snooker