# AI Integration Engineer

## Summary
- Organization: Angel7 AI pvt limited
- Location: Remote
- Type: Full-Time
- Department: Engineering
- Status: active
- Posted: [object Object]
- Updated: [object Object]
- Closing Date: N/A
- External Apply: No
- External Apply URL: support@talentreview.ai

## Details
- Salary: N/A
- Experience: N/A
- Education: N/A
- Team: N/A
- Reporting To: N/A

## Description
LLM Pipelines, RAG & Semantic Retrieval
Duration: 4 Weeks (Full-Time Only)
Location: Remote — India
Start: Immediate (within 1 week)

**About the Engagement**

We are a Singapore-based technology company building an enterprise AI automation product.

We are hiring an AI Integration Engineer for a focused 4-week sprint. You will own the product's intelligence layer — everything that connects language models, retrieval systems, and document processing to the user experience.

This is a full-time contract engagement (4 weeks), remote from anywhere in India. You will work within a 4-person sprint team.

This workstream benefits significantly from AI coding tools — strong usage of Cursor, Windsurf, or GitHub Copilot is expected.

**The Role**

- You are not building the AI model.
- You are building the system that:
- Routes tasks to the right model
- Manages inference cost
- Implements retrieval pipelines
- Processes documents for knowledge grounding
- Ensures reliability, observability, and performance
- You will design and implement the AI plumbing that makes the product intelligent and economically viable.

**What You Will Be Building**

**LLM Routing Layer**

- Unified adapter over multiple AI providers (Anthropic Claude, OpenAI) selecting models based on task type, cost, and capability.

**Cost Management System**

- Per-session and per-task spend tracking
- Configurable budget caps
- Escalation on breach

**Semantic Retrieval System (RAG)**

- Vector store implementation
- Embedding generation
- Similarity search
- Context assembly for grounded generation

**Document Ingestion Pipeline**

- Extraction → Chunking → Embedding → Storage → Retrieval

**Prompt Engineering Framework**

- Structured outputs (JSON)
- Tool use/function calling
- Few-shot examples
- Output validation


**Token Budget Management**

- Context window optimisation
- Rolling summaries
- Pre-inference cost estimation

**Required Skills (Must Have)**

- Anthropic Claude API (messages endpoint, tool use, structured outputs)
- OpenAI API (completions, function calling, embeddings)
- Production RAG pipeline experience (shipped to real users)
- Vector databases (ChromaDB, pgvector, Pinecone, Weaviate, etc.)
- Embedding models (e.g., text-embedding-3 or equivalent)
- Prompt engineering with structured JSON outputs
- Python (async, Pydantic, type hints)
- Token counting and cost estimation
- Document processing (PDF extraction, chunking strategies)

**Nice to Have**

- LangGraph (stateful orchestration)
- LangChain familiarity
- OCR pipelines (Tesseract or equivalent)
- Reranking models (Cohere Rerank or equivalent)
- Streaming LLM responses (SSE, async generators)
- LLM observability tools (LangSmith, Helicone, etc.)
- Knowledge graph concepts (Neo4j or similar)
- What Good Looks Like

**We are looking for someone who:**

- Has shipped a real RAG pipeline and can describe a production failure they debugged
- Can estimate inference cost before writing code
- Has used tool use/function calling beyond simple completions
- Iterates on prompts with measurable quality targets
- Understands embeddings deeply enough to explain retrieval errors
- Uses AI coding tools heavily to move fast
- This is a speed + systems thinking role.

**Engagement Details**

- Duration: 4 weeks (Full-time only — no part-time applicants)
- Location: Remote — India
- Hours: Full working day overlapping Singapore time
- Start: Immediate (within 1 week)
- AI Tooling: Cursor, Windsurf, or GitHub Copilot — heavy usage expected

## Responsibilities
None

## Skills
None

## Tags
None

## Organization
- Name: Angel7 AI pvt limited
- Website: N/A
- Industry: N/A
- Size: N/A
- Founded Year: N/A
- Description: N/A