Looking for a job

Smriti Tomar

@smrititomar

AI/ML Engineer at Parsewave.AI who cut data QA effort by 40% and improved downstream model accuracy by 95%.

At Parsewave.AI, I own the quality pipeline for multi-turn LLM training datasets, auditing and repairing data before production training. I developed the Divergent Test Ratio approach, which was adopted team-wide and reduced compute costs by about 50% while preserving training signal reliability.

At Intel, I built a Python pipeline to analyze firmware validation workflows, reducing manual debugging effort by 40%. My projects include a CrewAI multi-agent workflow, a PEFT pipeline for domain-specific NLP, and DocuMind AI, a RAG-based PDF question-answering system.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Data QC Engineer

Current

Parsewave.AI

Aug 2026 - Present (1 month)

Owned and managed end-to-end quality pipeline for 100+ multi-turn LLM training datasets, auditing, diagnosing, and repairing integrity issues before production training. Developed and standardized a statistical variance-evaluation approach (Divergent Test Ratio), reducing compute costs by ~50% while preserving training signal reliability.

LLM Data Quality Statistical Analysis Automation GPTZero Pangram

AI Engineer

Intel

Jun 2025 - May 2026 (11 months)

Built an LLM-assisted validation analysis pipeline in Python to automate interpretation of 120+ firmware validation workflows, reducing manual debugging effort by 40% and accelerating turnaround time by 25%. Leveraged Agentic AI workflows with intelligent log reasoning to analyze execution traces across 8 memory-domain features, increasing anomaly detection accuracy to over 92%.

Python LLM Agentic AI Firmware Validation Anomaly Detection Feature Engineering Automation