Looking for a job

Parth Jain

@parthjain

AI engineer building LLM agents and evaluation systems; improved CalorAI safety checks and cut destructive tool calls from 64% to 16%.

India

Message

On CalorAI, I collaborated with the founders on a LangGraph agent and built an outcome-based evaluation suite. In adversarial tests, I cut destructive wrong-tool calls from 64% to 16% and improved safety checks.

At PRAN Foundation, I implemented a 3D U-Net pipeline for electron-microscopy neuron segmentation. A data fix lifted Dice from 0.39 to 0.92, and mixed precision delivered about 78x speedup.

At ScaleAI (Outlier), I red-teamed frontier LLMs with adversarial tool-use tasks and created evaluation rubrics and supervised fine-tuning datasets. At Haltdos, I contributed to a no-code automation engine and developed ML systems that reduced manual intervention and fault-detection time.

Experience

Work history, roles, and key accomplishments

Research Intern, Connectomics and Computational Biology

PRAN Foundation

Mar 2026 - Aug 2026 (5 months)

Implemented a 3D U-Net deep learning pipeline for electron-microscopy neuron segmentation, achieving a ~78x speedup with mixed precision and improving Dice from 0.39 to 0.92. Optimized inference via roofline analysis and first-authored a review mapping 63 papers to 9 research gaps.

Deep Learning 3D UNet PyTorch Mixed Precision INT8 Quantization Roofline Analysis Connectomics Computational Biology Research Scientific Writing

AI Training & Evaluation Specialist

ScaleAI

Jan 2026 - Jun 2026 (5 months)

Red-teamed frontier LLMs with adversarial multi-step tool-use tasks, achieving 85%+ failure rates. Created supervised fine-tuning datasets and wrote 15+ criterion rubrics per task with zero QC defects.

Red Teaming LLM Evaluation Adversarial Testing Supervised FineTuning Rubric Design Quality Control GPT Claude

Machine Learning Intern, Custom RPA Core System

Haltdos

Jun 2025 - Aug 2025 (2 months)

Contributed to a no-code SOAR-style engine and engineered a gradient-boosted ML classifier for security workflows, reducing manual intervention by 85%. Created anomaly detection on distributed system logs, reducing mean fault detection time by 35%.

Machine Learning Gradient Boosting Python Anomaly Detection Feature Engineering Security Workflows SOAR Robotic Process Automation