Fourkites - Staff Software Engineer - AI/ML System

Description :

What you'll be doing :

Technical Leadership :

- Design and implement production-scale AI agent systems and orchestration frameworks (LangGraph, LangChain, similar architectures)

- Lead architecture for multi-agent systems handling complex business workflows

- Optimize deployment strategies using both LLMs and SLMs based on use case requirements

- Build natural language-configurable business process automation frameworks

- Implement multi-modal AI systems for document understanding (tables, charts, layouts)

AI/ML Implementation & Optimization :

- Deploy and optimize LLMs/SLMs in production with fine-tuning techniques (LoRA, QLoRA, DPO)

- Implement quantization strategies (INT8, INT4) and model distillation for edge deployment

- Build evaluation frameworks including LLM-as-judge systems and regression testing

- Design streaming architectures for real-time LLM responses (SSE, WebSockets)

- Create semantic caching and embedding-based retrieval systems

- Develop GraphRAG and long-context handling strategies (100k+ tokens)

System Architecture & Engineering :

- Design scalable microservices with comprehensive observability (LangSmith, Arize, custom telemetry)

- Build secure multi-tenant systems with prompt injection prevention and output validation

- Implement cost optimization through intelligent model routing and fallback strategies

- Develop document processing pipelines with OCR and layout understanding

- Create event-driven architectures for real-time shipment tracking and exception handling

Data & Infrastructure :

- Build data pipelines for training data curation, synthetic generation, and PII masking

- Implement RLHF/RLAIF feedback loops for continuous improvement

- Design experiment tracking and model registry systems (MLflow, DVC)

- Optimize inference costs through batch processing and spot instance utilization

- Establish model governance, audit trails, and compliance frameworks

About the team :

The AI team at FourKites builds the intelligence layer behind our next-generation AI workforce, powering autonomous agents used by Fortune 500 supply chains every day.

We operate at the intersection of production-grade AI, distributed systems, and real-world logistics, focusing on scalability, reliability, and measurable business impact.

The team values deep technical ownership, rapid experimentation, and thoughtful optimizationbalancing model performance, cost, and latency while building systems that operate at massive scale.

Who you are :

Technical Skills :

- 8+ years software engineering, 3+ years in production AI/ML systems

- Expertise in Python, PyTorch/JAX, and AI frameworks (LangChain, Transformers, PEFT)

- Experience with LLMs (GPT-4, Claude, Gemini) and SLMs (Phi, Llama, Mistral)

- Hands-on experience with :

1. Fine-tuning techniques (LoRA, QLoRA, DPO, RLHF)

2. Model optimization (quantization, distillation, pruning)

3. Vector databases and RAG architectures

4. Streaming systems and real-time processing

- Security measures (prompt injection prevention, jailbreak detection)

- Strong background in distributed systems, Kubernetes, and cloud platforms

Domain Knowledge(nice to have) :

- Experience with document intelligence and multi-modal AI systems

- Understanding of supply chain operations, EDI/API integrations

- Knowledge of token economics and consumption-based pricing models

- Familiarity with enterprise compliance requirements (GDPR, CCPA, SOC2)

Professional Skills :

- Track record of delivering complex projects with measurable business impact

- Experience with technical sales support, POCs, and customer success

- Strong communication for technical and non-technical audiences

- Data-driven decision making for model selection and cost optimization

Preferred Qualifications :

- Supply chain, logistics, or transportation management experience

- Experience with OCR pipelines and document extraction at scale

- Knowledge of GraphRAG and knowledge graph integration

- Contributions to open-source AI projects (Hugging Face, Ollama)

- Experience reducing inference costs by 50%+ through optimization

- Familiarity with MoE architectures and constitutional AI approaches

- Background in building usage-based billing and margin optimization

- Experience with specialized tools (vLLM, TGI, Triton, ONNX, TensorRT)

What You'll Work On :

- Building specialized AI agents solving supply chain problems

- Fine-tuning domain-specific models for supply chain terminology

- Implementing hybrid architectures combining cloud LLMs with edge SLMs

- Creating secure document intelligence systems for Fortune 500 clients

- Developing real-time exception handling for shipment tracking

- Building observability and evaluation frameworks for agent performance

- Designing fallback strategies and multi-provider redundancy

Technical Environment :

- Models : GPT-4, Claude, Gemini, Llama 3, Mistral, Phi-3, custom fine-tuned models

- Fine-tuning : LoRA/QLoRA, PEFT, DeepSpeed, bitsandbytes, Axolotl

- Infrastructure : Kubernetes, AWS SageMaker/Bedrock, GPU clusters, edge devices

- Frameworks : LangChain, LangGraph, vLLM, FastAPI, Transformers

- Observability : LangSmith, Weights & Biases, custom telemetry

- Data : PostgreSQL, Redis, Vector DBs, Kafka, feature stores

Impact & Growth :

- You'll directly contribute to AI initiatives generating millions in revenue while shaping systems processing millions of transactions daily.

- Lead technical decisions affecting 25+ engineers while mentoring the next generation of AI engineers.

- Be at the forefront of production AI optimization, balancing performance, cost, and latency for enterprise customers