Implementation Technical Lead

The Role You are an embedded builder who bridges frontier AI products and production-grade reality inside our clients’ operations. After our Fusion Discovery practice ships the prioritized roadmap, you are the named technical principal who turns that roadmap into running software — code, deploy, integrate, ship — inside the client’s Azure subscription, against their data, under their compliance constraints, on the calendar they signed up for. You are the FDE in the literal sense the term means in 2026: an embedded operator with a founder’s mindset who owns outcomes end to end. Discovery hands you a scoped milestone roadmap and a Statement of Work. You take it from there.You can expect to travel up to 30% of the time with this position. What This Role Is — And What It Is Not You will:write production code that ships into the client’s runtime. Build agentic systems withPydanticAI. Stand up event-driven ETL on Azure Container App Jobs. Scaffoldlovable.dev→ Static Web App frontends backed byFastAPIor ASP.NET Core BFFs (Backends-for-Frontends). Provision the whole stack with Bicep +azd. Wire structured logging, LLM cost telemetry, and the auth gateway from commit #1. Own the engagement’s technical outcome. You will not:run discovery interviews,facilitateexecutive prioritization workshops, or produce the Solution Architecture Document. Our Fusion Discovery practice does that upstream and handsyoutheir output. You may sit in on the final readout to take ownership of the implementation phase. The customer-facing volume sits with Discovery. You carry the technical volume. The Outcome We Hire You For In your first twelve months, you will serve as the technical lead on three to five Fusion Development engagements. Each one ships: A working production deployment in the client’s Azure subscription, deployed viaazdupand reproducible on demand. Vertical slices of demonstrable client value delivered against the milestone roadmap from Discovery. At least one KPI moved, with structured telemetry to prove it. A clean handoff document for ongoing operations: runbooks, infrastructure contracts, and the digital twin of the decisions made along the way. What You Will Actually Do Approximate time split: ~45% writing code, ~30% integration, infrastructure, and observability plumbing, ~15% client technical touchpoints (architecture standups, demo days, escalations), ~10% specifications, interface contracts, and engagement reflections.The customer-facing volume is lower than Big Lab FDE pool roles because PTM’s Discovery practice carries the customer-facing volume upstream. Convert the Discovery roadmap into an executable engineering plan withswimlanes, interface contracts, and validator-firsttodos. Scaffold the first commit as a working local full stack — frontend, backend, database, auth gateway, observability,IaC(Infrastructure as Code) — bootable with one command on a fresh clone. We do not believe in scaffold-then-wire-later. Implement Python services (FastAPIBFFs, event-driven ETL on Azure Container App Jobs,PydanticAIagentic systems) under our RPIR validator-first loop: Research, Plan,Implementvalidators RED, implement production code GREEN, Review. Build ASP.NET Core BFFs when the client’s existing investment or team strength calls for it. Use .NET AspireAppHostfor local orchestration when there are three or more services or any polyglot mix. Wire frontends scaffolded vialovable.devinto your BFF, hosted on Azure Static Web Apps. Stand up the Azure footprint with Bicep +azd: subscription-level deployments, module contracts, mandatory tagging, Key Vault-backed configuration, andwhat-ifgates on every pull request. Instrument structured logging from commit #1 (structlogwithcallsitemetadata), wire OTEL (OpenTelemetry) to the Aspire dev dashboard locally and to Azure Monitor inproduction, andemit LLM cost telemetry on every model call. Stand up the auth gateway:TraefikwithForwardAuthin containers, Microsoft Entra ID by default, with Auth0, Okta, Google Workspace OIDC, AWS Cognito, or generic OIDC as first-class alternates when the client’s identity posture calls for it. Pair-program with the second engineer on the engagement under Maker-Checker. Solo delivery is not a PTM pattern. Address the blockers that prevent AI from reaching enterprise-grade maturity in real environments: data readiness, integration complexity, state management, identity boundaries, observability gaps, and LLM cost runaway. Feed lessons back intoPTM’s livingdigital twin — the Fusion Covenant, the persona library, the agentic-coding template, the methodology. Every engagement you ship updates our rules. The Stack You Will Work In PTM is Azure-first and opinionated. The stack is documented. Deviations require architect approval with written rationale captured in the SAD. Backend:Python (FastAPI,Pydantic,PydanticAIfor agentic systems) and ASP.NET Core (latest LTS) for BFFs and APIs. Choose by client posture and team strength. Frontend:React + Vite scaffolded vialovable.dev, hosted on Azure Static Web Apps. Data:Azure SQL, Dataverse, Microsoft FabricOneLake, Azure AI Search. PostgreSQL when extensions justify it (Azure Flexible Server or self-hosted on Azure VM). Pure Python ETL withduckdb+delta-rsoverPySparkunless the client has a Fabric mandate. AI:Azure AI Foundry primary; OpenAI and Anthropic direct as alternates.PydanticAIfor agentic systems with typed state, structured outputs, and tool calls. 12-factoragentsprinciples. Infrastructure:Bicep +azd, GitHub Actions, Docker + Azure Container Registry. .NET AspireAppHostfor local orchestration whenwarranted. Observability:Azure Monitor, Log Analytics, ApplicationInsightsJS SDK on the frontend. Structured JSON logs with correlation IDs from day one. LLM cost telemetry mandatory on every model call. Identity:Entra ID default. Auth0, Okta, Google Workspace OIDC, AWS Cognito, and generic OIDC are first-class alternates.Traefik+ForwardAuthas the containerized gateway pattern. Local orchestration:.NET AspireAppHostwhenthere are three or more services, docker compose for homogeneous pairs,uvrunfor single-service spikes. One-off scripts:uvwith PEP 723 inline metadata. Never a half-builtrequirements.txt. What PTM’s Engineering Discipline Looks Like This is the part that differentiates PTM from a Big Lab FDE pool. We are opinionated about how engineering happens. Validators first, always.Tests, linters, and type-checkers are written andconfirmedRED before production code is written. Production code exists to turnvalidatorsGREEN. Python projects mandateruff+pyrighton every commit. Contracts before code.Every interface — API boundary, DB schema, shared type, inter-service message — gets aformal.contract.mdin/docs/schema/during planning. Implementation does not begin until the contract is accepted. Right-sized process.Trivial fixes go direct. Medium work uses the RPIR loop. Large work decomposes intoswimlaneswith parallel specialists. We do not skipprocessfor big work, and we do not over-process small work. First commit is a working local full stack.Frontend, backend, database, auth gateway, observability,IaC, bootable with one command on a fresh clone. No “scaffold first, wire later.” This is the floor, not the ceiling. Maker-Checker over solo delivery.Pair on architectural choices, contracts, and material deliverables. Living digital twin.When reality drifts from specification, we update the specification. Every engagement-significant lesson is captured in a reflection and folded back into themethodology. If this sounds like overhead, this is not the right firm. If this sounds like the engineering culture you have been missing on FDE pool work, keep reading. What “Senior” and “High-Agency” Mean Here Same as the FDE category at large, with PTM’s specific edges: You can stand up a working greenfield Azure deployment from a Discovery roadmap, solo, in under a week of focused work. You can read a Discovery Solution Architecture Document andidentifythe three implementation risks the consultant did not see. You raise them in the engagement kickoff, not in production. You can write a contract for an interface that has not been builtyetand have itbeuseful to a second engineer six weeks later. You can sit in a client architecture standup and defend a respectful “no” to apoorly-consideredscopeaddition without losing the relationship. You own outcomes. The engagement either ships itsmilestonesor it does not, and you carry that. PTM gives you scope (via the Discovery handoff),methodology(the Fusion Covenant, the agentic-coding template, the persona library), and a second engineer for Maker-Checker pairing. Within that, youoperatewith high autonomy. Qualifications Seven or more years of production software engineering experience, with at least three at senior-IC or tech-lead level. Deep Python: typed code, async, packaging viauv,FastAPI,Pydantic.PydanticAIor comparableagentic-systemsexperience is a strong plus. Working competence in TypeScript + React. You do not need to be a frontend specialist; you do need to be able to wire alovable.devscaffold into your BFF without help. Azure production experience: Container Apps, Functions, Static Web Apps, Azure SQL, Key Vault, Bicep +azd. Equivalent depth in another cloud is acceptable if you are willing to convert quickly. Observability fluency: structured logging, distributed tracing, OTEL, log-trace correlation. You have shipped a service where you knew what was happening in production because you instrumented it that way. Test discipline: you have written tests before production code on purpose, and you can explain why.pytest+ property-based testing experience preferred. Comfort with client-facing technical work: architecture standups, escalations, demo days. You do not need to be a presenter; you do need to be able to defend a technical decision in a room that includes the client’s CTO. US workauthorization. Nice to Have ASP.NET Core / C# production experience. Healthcare engagement experience, particularly in HIPAA-aware system design, EHR integration, FHIR / HL7, claims data, or revenue-cycle-adjacent systems. The first twelve months of engagements arehealthcare-weighted. Experience withdotenvx, SOPS, or comparableencrypted-configurationworkflows. Experience with .NET AspireAppHostor a comparable multi-service local orchestration tool. PriorFDE, solutions-architect, or embedded-consultant experience at a vendor or services firm. How This Role Differs from FDE Roles at the Big Labs You are watching the samemarketwe are. Google Cloud is hiring FDEs aggressively under a shortened two-day interview cycle. The OpenAI Deployment Company is hiringhundredsinside a quasi-external entity capitalized by TPG and Advent. Anthropic is standingupa similar venture with Blackstone, Hellman & Friedman, and Goldman Sachs. Those roles are real, large, and visible. PTM is the alternative for senior engineers who would rather: Be a named principal on an engagement, with a documented Discovery handoff defining scope, instead of one FDE in a hundred-person pool inventing scope on the customer’s whiteboard. Work inside a documented engineering discipline (validators-first, contracts-first, RPIR, Maker-Checker) instead of “high-agency” being a euphemism for “no resources, no spec, your problem.” See your engagement lessons land in the firm’smethodologywithin weeks, instead of filed as tickets that may or may not be read. Ship working systems on a documented stack instead of “white-glove” delivering whatever the customer’s loudest stakeholder requested last week. Why This Role Exists Now PTM’s Fusion Development practice is scaling. The Discovery practice (see the companion role, Forward Deployed Engineer — AI Consultant) is generating fundable roadmaps faster than we can implement them. You will be one of the first technical leads on the Implementation side. Themethodology, the stack, theIaCtemplates, the agentic-coding patterns, and the engagement pipeline are all real today. What is not yet built is the bench. That istheopportunity. Possibility of contract-to-hire Full Medical Benefits 2 Weeks Paid Vacation Full Time Dental & vision insurance 401(k) matching #J-18808-Ljbffr Proactive