Back to ConForge AI
Open roles
AI Engineer Remote, United States · Full time
About the role
You own the architecture, write the code, take it to production, and lead our AI initiatives as the engineering team grows. Software engineering and applied AI in equal measure.
5+ years of engineering experience, including time in an early-stage environment where nothing was built yet.
Key responsibilities
Architecture
- Own the end-to-end system design: platform, web, mobile, data, security
- Make build, buy and integrate decisions across every layer
- Define the technical standards the team works to
Data and retrieval
- Build ingestion pipelines for structured and unstructured enterprise data
- Design hybrid retrieval across vector, keyword and metadata search
- Implement knowledge graphs and persistent memory
Documents and vision
- Build parsing pipelines for complex technical documents at scale
- Handle real-world input: scanned, rotated, marked-up, inconsistent
- Apply OCR and computer vision where text extraction is not enough
AI systems
- Design multi-agent architectures with supervisor and sub-agent patterns
- Build orchestration, tool calling, state management and human-in-the-loop
- Own fine-tuning pipelines, adapter versioning and model routing
Evaluation and observability
- Build the eval harness before any fine-tuning begins
- Implement tracing, cost and latency tracking, drift and regression alerts
- Own accuracy, grounding and hallucination detection
Production
- Take the platform to production and keep it there
- Own infrastructure, deployment, CI/CD, monitoring and security
- Support cloud, on-premise and air-gapped deployment
Delivery and team
- Translate the product roadmap into working software
- Work directly with the people who will use this, and turn how they work into how the system works
- Set delivery cadence, quality gates and engineering standards
- Hire, onboard and lead the engineering team
Technologies
You will not have used all of these. You should be strong across most, and able to pick up the rest.
| Languages and backend | Python, TypeScript, SQL, FastAPI, Node.js, REST, gRPC, async services |
| Generative AI | LangChain, LangGraph, MCP, A2A, Claude, OpenAI, open-source models (Llama, Mistral, Qwen) |
| Agents and orchestration | Multi-agent supervisor patterns, task decomposition, tool and function calling, parallel sub-agents, state management, human-in-the-loop, agent and skill registries, CrewAI, Google ADK |
| Workflow orchestration | Airflow, Celery, event-driven pipelines, scheduled jobs, retry and fallback logic |
| Retrieval | Vector databases, embeddings, hybrid and semantic search, GraphRAG, knowledge graphs |
| Documents, OCR and vision | PDF and technical document parsing, OCR, layout and table extraction, image labelling, computer vision, unstructured data pipelines |
| Fine-tuning | LoRA, QLoRA, PEFT, adapter versioning, model distillation |
| ML and data | PyTorch, Spark, Kafka, Delta Lake, ETL and streaming pipelines |
| MLOps and evaluation | MLflow, LangFuse, LangSmith, W&B, Portkey, experiment tracking, eval harnesses, LLM-as-judge, drift detection, model routing and fallback |
| Frontend and apps | Next.js, React, React Native |
| Databases | PostgreSQL, MongoDB, Redis, Elasticsearch, Neo4j |
| Cloud and infrastructure | AWS, Azure or GCP, Docker, Kubernetes, Helm, Terraform, GitHub Actions, CI/CD |
| Observability | Prometheus, Grafana, OpenTelemetry |
Required experience
- Built and shipped an enterprise-grade AI or platform product end to end in production
- Deep hands-on experience with agent frameworks, orchestration and tool calling
- RAG in production: vector search, embeddings, hybrid retrieval, evaluation
- Document processing at scale, including messy and unstructured real-world input
- Fine-tuning and model operations: dataset curation, training, benchmarking
- Strong backend, API, data and cloud engineering. Hands on, not architecture only
- Containers, orchestration, CI/CD, observability and security
- Evaluation as a discipline: eval sets, LLM-as-judge, hallucination detection, drift
Preferred
- Founding or first engineering hire at an early-stage company
- Experience building from nothing rather than inheriting an existing system
- Computer vision or advanced OCR pipeline experience
- Customer-facing technical background, such as forward deployed or solutions engineering
- Track record shipping under real business constraints and fixed timelines
- Comfortable with the reality of early stage: no playbook, no safety net, and the pager is yours
What success looks like
- The platform is live, in production, with real users
- Releases ship on cadence and quality holds
- The evaluation framework tells us whether each change helped
- The engineering team grows and does its best work