AI Builder

Mekari; Jakarta Pusat; Sur place; IDR 60 000 000 - 100 000 000; Plein temps

Résumé du poste

Mekari seeks a Senior Platform Engineer to build and operate secure, observable inference APIs and routing policies in Jakarta. You will deliver versioned agent drivers, task-specific models, and retrieval pipelines for product integration.

You will work on evolving governance, cost controls, and observability dashboards while ensuring scalable, safe deployments across platforms.

Qualifications

• Experience designing scalable APIs for high-concurrency workloads.
• Knowledge of authentication, authz, and security best practices.
• Ability to work with cross-functional teams across platforms.

Responsabilités

• API & Platform Architecture: Design high-concurrency REST/streaming APIs, handle auth (OAuth2), RBAC, tenant isolation, fallbacks, metering, and PostgreSQL data modeling.
• Model & Agent Delivery: Integrate provider/self-hosted models, maintain versioned capability catalogues, and build lightweight, task-specific models and agent drivers.
• Routing & Optimization: Run policy-aware routing, run benchmarks across prompt/output variations, and optimize throughput, latency, and token economics.
• Evaluation & Safety: Partner with experts to curate task datasets, set release thresholds, run baseline comparisons, and deploy shadow/staged releases safely.
• Operations & Governance: Maintain observability dashboards, alerts, runbooks, cost controls, tenant data privacy, clear technical documentation, and code quality.

Connaissances

REST/Streaming API design

OAuth2 authentication

RBAC & tenant isolation

Observability & monitoring

PostgreSQL data modeling

Outils

PostgreSQL

OAuth2

RBAC

Prometheus

Grafana

Description du poste

• Platform Shipping: Build and run secure, observable inference APIs, routing policies, tenant controls, metering, and operator control surfaces.
• Intelligence Supply: Deliver versioned agent drivers, task‑specific small models, prompts, tools, and retrieval pipelines for product integration.
• Evaluation & Feedback: Use production data, benchmarks, quality gates, and cost metrics to continuously improve models, routing, and guardrails.

How The Role Ships End-to-End

• Platform Shipping: Build and run secure, observable inference APIs, routing policies, tenant controls, metering, and operator control surfaces.
• Intelligence Supply: Deliver versioned agent drivers, task‑specific small models, prompts, tools, and retrieval pipelines for product integration.
• Evaluation & Feedback: Use production data, benchmarks, quality gates, and cost metrics to continuously improve models, routing, and guardrails.

Core Responsibilities

• API & Platform Architecture: Design high-concurrency REST/streaming APIs, handle auth (OAuth2), RBAC, tenant isolation, fallbacks, metering, and PostgreSQL data modeling.
• Model & Agent Delivery: Integrate provider/self-hosted models, maintain versioned capability catalogues, and build lightweight, task‑specific models and agent drivers.
• Routing & Optimization: Run policy‑aware routing, run benchmarks across prompt/output variations, and optimize throughput, latency, and token economics.
• Evaluation & Safety: Partner with experts to curate task datasets, set release thresholds, run baseline comparisons, and deploy shadow/staged releases safely.
• Operations & Governance: Maintain observability dashboards, alerts, runbooks, cost controls, tenant data privacy, clear technical documentation, and code quality.

Typical Deliverables

• Production inference APIs, web control surfaces, tenant quota controls, and accurate usage metering.
• Versioned model integrations, fallback execution rules, capacity plans, and performance benchmarks.
• Task‑specific model endpoints and agent drivers with clear contracts, owners, and versioning.
• Evaluation datasets, release gates, regression checks, and observability runbooks.
• Technical design notes detailing model selection, routing logic, trade‑offs, and unit economics.

Success Measures

• Adoption & Usability: Growing adoption across product teams through a single governed, reusable platform path.
• Quality & Safety: Higher task success rates alongside strict adherence to safety, quality, and tenant policy boundaries.
• Economic & Performance Efficiency: Lower cost and latency per completed task, driven by smaller models and optimized routing.
• Reliability & Agility: Reversible, progressive releases with early detection of regressions, anomalies, or capacity bottlenecks.