LogoFresh Jobs Ahead
FourKites, Inc.
FourKites, Inc.
501 – 1,000 Employees
Food & BeverageLogisticsManufacturing
FourKites, Inc. develops a global supply chain visibility platform that helps organizations monitor and coordinate logistics in real time. Its applications bring together data across transportation, inventory, and yard operations, giving teams a clearer view of shipments and supply chain activity across multiple modes of transport. The company serves more than 1,500 major brands, including businesses in logistics, food and beverage, pharmaceuticals, and manufacturing, and connects with enterprise systems to support more efficient operations, cost control, and customer service.
LinkedIn
Company Jobs

Principal AI Architect, Agent Systems

Lead the architecture for FourKites’ AI agents, from models and context management to evaluation. Build the platform behind AI-powered supply chain digital workers.

INIndiaHybridFull-timeSeniorLeadEnglish RequiredAI EngineerMachine Learning Engineer
AWSGoogle Cloud PlatformJavaKafkaKubernetesMicroservicesPostgresPythonTypeScript

Description

  • Own the architecture for FourKites AI agents, covering execution, context, models, and evaluation.
  • Set the agent reference architecture for loops, tool use, state, checkpoints, retries, recovery, safeguards, permissions, and human handoffs.
  • Design durable, multi-step workflows that cross context windows and recover from tool, carrier API, and model failures.
  • Choose the right approach for each problem, from frontier or open-weight models to fine-tuned encoders or conventional code.
  • Manage model routing, per-task costs, and latency targets across the agent fleet.
  • Connect agents to Kafka, microservices, Postgres, customer APIs, and MCP servers.
  • Define context-window standards for prompts, tools, retrieved content, memory, and prior turns within token limits.
  • Create shared tools for retrieval, context compaction, summarization, sub-agent isolation, prompt caching, and persistent memory.
  • Investigate context failures and turn solutions into reusable execution patterns.
  • Develop offline evaluation systems using production traces, code-based checks, calibrated LLM-as-judge grading, and CI regression suites.
  • Develop online evaluation with production scoring, quality measures, drift monitoring, shadow and A/B releases, feedback signals, and quality alerts.
  • Link evaluation measures to outcomes such as tracking accuracy, automated exception resolution, and appointment fill rates.
  • Review designs and pull requests, document architecture and operating playbooks, and mentor staff and senior engineers.
  • Follow model and tooling developments, run experiments, and recommend whether to build, buy, or wait.
  • Work with Product, Solutions, and Customer Success to turn field issues into platform improvements.

Requirements

  • Have architected and operated an LLM agent system serving real users and production traffic, including tool calls, multi-step workflows, failure handling, and escalation to people.
  • Have owned an LLM agent system after launch, managing incidents, cost, and latency.
  • Bring deep machine learning knowledge, including attention, positional encodings, encoder and decoder architectures, tokenization, and KV caches.
  • Have trained or fine-tuned encoder models, such as BERT-family models, sentence embeddings, bi-encoders, and cross-encoder re-rankers.
  • Understand contrastive training, distillation, and quantization.
  • Have managed token budgets and built retrieval and memory for agents.
  • Have used context compaction and sub-agents, and measured quality as context windows fill.
  • Have built agent execution layers with loops, tool dispatch, state and checkpoints, interruption and recovery, sandboxing and permissions, safeguards, and model routing.
  • Be fluent in MCP, structured outputs, and streaming.
  • Have built offline and online evaluation suites covering datasets, graders, LLM-as-judge calibration, CI regression gates, production scoring, trace sampling, drift monitoring, shadow deployments, and feedback loops.
  • Bring 10+ years of software or machine learning systems experience, including 3+ years building with LLMs.
  • Have strong, hands-on Python skills.
  • Be comfortable reading and reviewing Java or TypeScript services.
  • Have production experience with AWS or GCP, containers, Kubernetes, and event-driven systems.
  • Communicate effectively in writing and in conversation.
  • Preferred: experience in supply chain, logistics, or transportation.
  • Preferred: experience with Document AI, OCR, layout models, and extracting data from bills of lading, invoices, and proof-of-delivery documents.
  • Preferred: experience building voice or email agents that act for a business.
  • Preferred: experience with agent security, including prompt-injection defense, tool permissions, and tenant data isolation.
  • Preferred: experience with Claude, GPT, Gemini, open-weight models, and model-serving stacks.
  • Preferred: publications, talks, or open-source contributions on agents, retrieval, or evaluation.

Benefits

  • Competitive compensation and stock options.
  • Five global recharge days.
  • Generous paid time off and standard holidays.
  • Parental leave available to all parents.
  • Annual wellness stipend.
  • Volunteer days.
  • Medical benefits begin on the first day of employment.
  • 36 PTO days across sick, casual, and earned leave.
  • Five recharge days.
  • Two volunteer days.
  • Home office setup and technology reimbursement.
  • Lifestyle and family benefits.
  • Mental wellness support and guidance.
  • Ongoing learning and development, including a professional development program and Toastmasters club.

Related Jobs