Looking for a job

Divakar Mekala

@divakarmekala

Message

AI Engineer building SAIRE, a 20-stage LLM verification platform with governed decisions and audit traces.

India

Message

What I'm looking for

I'm seeking entry-level roles in AI engineering, LLM and AI agent development, AI training and evaluation, prompt engineering, or Python and full-stack development.

I built SAIRE, an enterprise AI control and reliability platform whose Phase 1 pilot runs LLM responses through a 20-stage verification pipeline. It produces governed decisions—ALLOW, WARN, ASK_HUMAN, or BLOCK—and stores an audit trace for each decision.

At Outlier, I evaluate LLM outputs for factual accuracy, reasoning quality, safety, and instruction adherence. I also design benchmark tasks for Project Dynamo and adversarial prompts for Project Seal India, while building and deploying projects including Agent Action Review and Truth or Trap.

Experience

Work history, roles, and key accomplishments

Current

AI Training Contractor

Current

Project Dynamo

Jun 2026 - Present (5 months)

Design structured evaluation tasks for benchmarking LLM performance in terminal and CLI reasoning, tool use, and multi-step agentic task completion. Create reproducible benchmark scenarios and review task quality and model behavior.

Terminal Bench Benchmark Task Design LLM Evaluation Agentic Workflows CLI Reasoning Tool Use

Current

AI Training Contractor

Current

Project Seal India

Jun 2026 - Present (5 months)

Write adversarial prompts to stress-test LLM web search and reasoning, exposing model failure modes. Evaluate search-grounded responses for citation accuracy, logical consistency, and edge cases.

Adversarial Prompt Design LLM Evaluation Citation Verification Reasoning AI Safety

Current

AI Model Evaluator

Current

Outlier

Jan 2024 - Present (2 years 10 months)

Evaluate LLM outputs for factual accuracy, reasoning quality, safety, and instruction adherence across diverse prompt categories. Detect hallucinations and validate responses to surface model weaknesses and failure modes.

LLM Evaluation Hallucination Detection Factual Accuracy Reasoning Safety Instruction Adherence