Intelligence Systems, Retrieval, and Workflow Architecture

Why This Role Exists

RRapid Alpha is transitioning from analyst-driven delivery to a platform-driven execution model.

Our goal is simple:

Multiply analyst output 10x while increasing quality, repeatability, and margin

We are building EVOS into a system that:

• Ingests thousands of documents
• Converts unstructured data into structured intelligence
• Generates scenarios and roadmaps
• Produces outputs that are traceable, auditable, and repeatable

This role exists to:

Design and implement the intelligence systems that make this possible

---

This role IS:

An architecture + implementation role responsible for building production AI systems that replace analyst work

---

What You Will Own

1. Intelligence System Design (Core)

You will design how EVOS produces intelligence.

This includes:

• classification systems
• retrieval systems
• reasoning workflows
• structured output generation

You will define:

• how evidence becomes structured data
• how systems generate scenarios and roadmaps
• how outputs remain auditable and reproducible

---

2. Document Ingestion & Transformation Pipelines

You will build systems that:

• process 5,000–50,000+ documents
• convert unstructured → structured data
• implement chunking strategies (semantic + hierarchical)
• perform multi-stage classification

---

3. Retrieval Systems (RAG Done Correctly)

You will design:

• hybrid retrieval systems (vector + metadata)
• ranking and filtering logic
• systems that minimize token usage and API calls

---

4. Agentic & Workflow Systems

You will design multi-step systems such as:

Evidence → STU → Scenario → Roadmap

This includes:

• orchestrating multi-stage workflows
• designing stateful processing
• ensuring outputs are structured and reusable

---

5. Human-in-the-Loop Systems

You will define:

• where analysts intervene
• how corrections are captured
• how feedback improves system performance

---

6. Evaluation & Accuracy Systems

You will define:

• precision / recall metrics
• ranking evaluation frameworks
• false positive / false negative analysis
• dataset-based testing

---

👉 This is critical:

Outputs must be correct, not just generated

---

7. Production-Grade System Implementation

You will build systems that:

• are asynchronous and queue-driven
• scale across clients
• integrate with platform infrastructure (AWS, Postgres, services)
• operate reliably under load

You will work closely with the Platform Architect to ensure:

• workload isolation
• system stability
• predictable performance

---

What You Must Be Able To Explain Clearly

You should be able to answer:

• How to process and classify 5,000+ documents across multiple dimensions
• How to design retrieval systems that minimize cost and maximize accuracy
• What fails first in LLM systems at scale
• How to evaluate system accuracy rigorously
• How to design workflows that produce auditable outputs
• Where human intervention is required in AI systems

---

Required Experience

• 6–10 years engineering experience
• 2+ years building production AI / NLP / search systems
• Strong Python
• Experience with:
  • vector databases (PGVector, Pinecone, Weaviate)
  • embeddings and retrieval systems
  • processing large document datasets
• Experience designing multi-step workflows or pipelines

• Built RAG systems in production
• Designed classification systems (not just used APIs)
• Worked with large document datasets
• Defined evaluation or ranking systems
• Built systems that combine automation with human review

---

Application Requirements

To be considered, you MUST include:

1. System Description

Describe one system you personally designed that:

• processes large document datasets
• performs classification, retrieval, or transformation
• produces structured outputs

2. Architecture Exercise

Explain:

How you would design a system to process 5,000+ documents and classify them across multiple dimensions, including ingestion, classification, retrieval, and evaluation.

Applications without this will not be reviewed.

---

What Success Looks Like

In the first 90 days:

• A working ingestion + classification pipeline
• Initial retrieval system operational
• Structured outputs generated from real data
• Evaluation framework defined

In 6 months:

• Analysts are significantly more productive
• Workflows are partially automated
• Outputs are consistent and auditable

In 12 months:

EVOS replaces large portions of analyst work with scalable systems

Application Requirements

To be considered, you MUST include:

1. A description of one production AI/LLM system you designed
2. A short explanation (max 500 words): How you would architect a system to process 5,000+ documents and classify them across multiple dimensions

Applications without this will not be reviewed.

Compensation & Working Model

• 100% Remote
  Work from anywhere. We operate as a distributed team and optimize for output, not location.
• India Federal Holidays + Company Shutdown Weeks
  In addition to standard holidays, we operate two company-wide shutdown periods:
  • June 29 – July 4
  • December 23 – January 6
    These are built-in recovery periods to support sustained high performance.
• Full Benefits
  Comprehensive benefits package aligned with senior engineering roles.
• High Ownership, High Accountability
  This is not a task-based role. You will own core systems that directly impact platform scalability, reliability, and company performance.
• Direct Collaboration with Founder
  You will work closely with the CEO and senior strategists to translate expert reasoning into production systems.

How to Apply

Send:

1. A brief description of one production AI/LLM system you personally designed.
2. A short explanation (max 500 words) of how you would architect a system to process 5,000+ research documents and classify them across multiple dimensions.
3. Your CV and expected CTC.

Applications without this information will not be considered.