SHREY MATHUR

About

Applied ML/LLM engineer currently building AppleCare’s production chat recommender, owning the full lifecycle from transcript cleaning and fine-tuning through distributed deployment, monitoring, and iterative improvement. Their work delivered >10% copy-count improvement, 5% lower modification rate, 60% cost reduction, and $1.1M profitability in 2025, and they also created a reasoning-data generation approach that enabled a reasoning model and a judge model that cut eval time by over 99%.

Key Strengths

• Owned production LLM system end-to-end from data prep through deployment and monitoring
• Strong applied LLM optimization across SFT, RLHF-style methods, PEFT, pruning, and MoE
• Research-to-production execution on novel clustering/sampling techniques
• Improved AppleCare advisor adoption and reduced modification rates with measurable business impact
• Drove 60% cost reduction and $1.1M project profitability in CY2025
• Built scalable low-latency inference using vLLM for SLA-sensitive production traffic
• Created synthetic reasoning dataset and shipped reasoning-enabled LLM feature
• Reduced evaluation time by 99%+ via judge model aligned to human evaluators
• Strong production judgment on latency-quality tradeoffs informed by advisor behavior
• Experience leading multilingual expansion after English-language success
• Automated manual reporting workflow at RBS, cutting turnaround from 2 hours to seconds

Experience

Education

Awards

• Reward&Recognition award - Royal Bank of Scotland
• Pre-Placement job offer - Royal Bank of Scotland

Skills

Languages