Alexander LaGonterie

Insurance-technology SaaS (17,000+ customers, 500,000+ users); AMS360 is its flagship agency-management platform (G2's top-rated AMS, 2025).

• Team Lead of a 7-person scrum team (5 engineers, QA, product owner) while remaining its senior IC: cut new-developer onboarding from days to minutes (~99%) with setup automation, and mentored engineers to autonomy — bringing an SE1 to independent operation — through daily code review and pair programming.
• Conceived and built Auth Engine (self-initiated), a Rust in-memory authorization engine (gRPC, Roaring Bitmaps) replacing AMS360's per-request SQL security resolution. Mirrors the legacy stored-procedure semantics exactly, with three-layer sync, zero-downtime atomic index swaps, graceful SQL fallback, and 600+ tests.
• On real 6.85M-policy production-scale data, Auth Engine resolves security-filtered searches and access-set queries — work that takes the SQL it replaces seconds to minutes — in microseconds: every customer a user can access in 26µs vs 455s (17,000,000×), filtered customer search in 146µs vs 439s (3,000,000×), in under 1GB of memory.
• Architected TMS, the .NET 10 successor to AMS360's two-decade-old backend, from a vague VP mandate — 3–5× faster new-API development — by re-architecting away from the legacy 3-layer design that had buckled as the codebase grew massive. Delivered the first domain-driven design AMS360 ever truly adopted (long discussed, never realized), enabled by the modern patterns that finally make DDD practical: CQRS with hybrid event sourcing over the live legacy database (new event types need zero schema changes) and single-round-trip transactional commits.
• Enforced the TMS architecture at compile time with 4 Roslyn source generators + 25 custom analyzers (zero runtime reflection), multi-tenant JWT/OIDC security with per-agency database routing, and a delivery platform of ephemeral per-feature-branch AWS environments (GitLab CI, AWS CDK) with OpenTelemetry into Dynatrace — now the reference blueprint for Vertafore's next-generation .NET services.
• Helped ship Vertafore's flagship customer-facing AI product, the Velocity AI Reconciliation Agent, as one of three selected AMS360 engineers integrating a new Python agent platform (FastAPI, LangChain, Postgres, multi-model via Portkey) into the legacy AMS360 system — contributing most directly on the AMS360 side across the data model, APIs, and integration architecture, and implementing the per-carrier "postback" learning loop that captures users' manual corrections and replays them on each carrier's next statement to compound accuracy. The launched product reports 90% time saved at 94% accuracy across 200+ carriers.
• Created Vertafore's AI-native developer tooling from scratch, unprompted: six dual-surface CLI + MCP tools giving engineers' AI agents safe, structured access to the daily dev stack — SQL Server (execution-plan-aware analysis; AI-guided SQL tracing for legacy-UI modernization), GitLab, Rally, Teams, an umbrella installer, and a shared core library — plus a Claude Code plugin marketplace (CI-managed) of composable agents, skills, and slash-commands that chain the tools into multi-step workflows: a story-to-MR skill, for instance, drives a Rally story through to a GitLab merge request. The tooling includes agent-recoverable error contracts, vendored/pinned binaries, self-update infrastructure; adopted by 20+ engineers and growing.
• Drove spec-driven development and agent-readable docs across the team's repositories — the foundation that turns precise, machine-readable specs into reliable agent output.
• Earlier (SE I/II): as a Warden for the AMS360 backend (a required code reviewer across its APIs and a standing voice in its conventions and architecture) and a trained Scrum Master for several PIs — reworked multitenant database versioning strategy (−20k+ lines, ~15% lower per-request overhead), parallelized cross-service report generation (50% faster), built a Python SQL-to-ORM schema tool (800+ legacy tables mapped in seconds), and won the company hackathon (CTO-judged, 30+ teams) with the API-observability project that preceded company-wide Dynatrace adoption.