AI/ML Engineer

Location: DHA Phase 4, Lahore, Pakistan (Onsite)

Timings: 5pm–2am PKT (Full-time)

Experience: ~2 years

About the Role

We are looking for a hands-on AI/ML Engineer who can take architecture and solution designs and actually build them, writing code, wiring up agents, integrating AWS AI services, and shipping working systems.

This role is NOT about designing scope or strategy. It's about execution: turning specs, diagrams, and use cases into functioning agentic pipelines.

What You will Do

Agent Development

• Build and deploy multi-agent workflows (task planning, reasoning, execution, and validation/guardrail agents) based on designs handed off by the architecture team
• Implement tool-calling, function-calling, and structured outputs for LLM agents
• Work with agent frameworks such as LangGraph / LangChain (or equivalent) to implement orchestration logic

AI Platform & Model Integration

• Build with LLM platforms such as Amazon Bedrock, OpenAI, or Mistral — model invocation, prompt orchestration, and guardrails
• Implement agent orchestration logic (e.g., using Amazon Nova, or framework-based orchestration like LangGraph)
• Write and deploy serverless functions (AWS Lambda or equivalent) for event-driven agent execution and tool calling
• Integrate OCR/document-understanding tools for ingestion and structured data extraction — e.g., Mistral OCR, Amazon Textract, or similar, based on cost/accuracy trade-offs
• Build knowledge grounding and retrieval pipelines (vectorization, embeddings, RAG) using tools like Bedrock Knowledge Bases, open-source vector DBs, or equivalent
• Compare and combine models across providers (OpenAI, Bedrock, Mistral, etc.) to pick the right tool for cost, latency, and accuracy, not locked into a single vendor

Integration & Data Work

• Build API integrations, database queries, and internal tool connectors for agents to call
• Implement RAG pipelines (retrieval-augmented generation) and agentic retrieval flows
• Handle prompt versioning, short-term/long-term memory implementation, and context management in code

Quality, Testing & Iteration

• Write tests and evaluation scripts to catch hallucinations, drift, and failure modes
• Debug agent behavior, trace execution logs, and iterate on prompts/logic based on real output
• Implement basic guardrails, logging, and cost-tracking as instructed by governance guidelines

Non-Negotiable (This Matters More Than Any Specific Tool)

• We will prioritize strong fundamentals and real problem-solving ability over exposure to specific frameworks or buzzwords. Specifically, you must have:
• Rock-solid programming fundamentals — data structures, algorithms, clean code, debugging skills that hold up under pressure
• Genuine problem-solving ability — can break down an ambiguous, half-defined problem into logical steps without being told exactly what to do
• Fast, structured learning ability — can pick up a new framework, API, or AWS service in days, not weeks, because the fundamentals are solid
• First-principles thinking — when something breaks, can reason from how it actually works rather than guessing or copy-pasting fixes
• Ownership mentality — doesn't stop at "it runs," pushes until it actually works correctly and handles edge cases
• Candidates who are strong on fundamentals but light on AI-specific experience are preferred over candidates who only know a specific framework tutorial-deep. We can teach Bedrock, LangGraph, or Nova. We cannot teach how to think.

What You Bring (Required)

• ~2 years of hands-on software development experience
• Strong Python skills, with genuine understanding of the language (not just syntax)
• Solid grasp of core CS fundamentals: data structures, algorithms, complexity, debugging methodology
• Some practical exposure to LLM APIs (OpenAI, Anthropic, Bedrock, or similar) — depth less important than proof you can learn this space quickly
• Working knowledge of at least one major cloud platform (AWS preferred: Lambda, IAM, S3) — GCP/Azure equivalents also fine
• Comfortable working with APIs, REST integrations, and structured data (JSON, databases)
• Familiarity with backend development — building/consuming REST APIs, working with databases, understanding request/response lifecycles, basic auth, and service-to-service communication (e.g., FastAPI, Flask, Node/Express, or similar)
• Debugging mindset — comfortable reading logs, tracing failures, and fixing broken behavior systematically rather than by trial and error
• Ability to take a spec/diagram from a senior architect and turn it into working code without hand-holding
• Bonus, not required: experience with an agent framework (LangChain, LangGraph, CrewAI, etc.) — we'd rather hire strong fundamentals and teach this than hire framework familiarity without the fundamentals

Nice to Have

• Experience with vector databases (Pinecone, OpenSearch, FAISS, etc.)
• Exposure to OCR/document-AI tools (Mistral OCR, Amazon Textract, or similar)
• Familiarity with RAG pipeline construction
• Experience in a startup or fast-shipping environment
• Basic understanding of prompt engineering best practices

What This Role Is NOT

• Not a solution architecture or client-facing consulting role
• Not responsible for scoping, SoW creation, or high-level roadmap decisions
• Not a "prompt-only" role — this requires real coding ability