Parth Ahuja

Summary

Integrated MikeLegal's IP Solutions into Litigator, centralizing trademark and patent case data within the litigation workflow, resulting in a 15-20% boost in paralegal research efficiency (based on internal benchmarks). MikeLitigator: Led the design and implementation of modules for client management, budgeting, expense tracking, timesheet logging, and subscription handling-streamlining internal operations and improving billing accuracy by 5% across 120+ active clients.

Summary

Attained 20% growth in client retention over one year by analyzing business needs, collaborating with stakeholders, and implementing applications. Led a 4-member team to develop an Internal Management Application for quarterly review, reducing review time by 83% and saving 1800 person-hours over 4 quarters. Sequence: Architected and implemented a domain-driven design for AdoptSequence, a platform that automates voice replacement with Al-generated voice-overs, resulting in a 30% reduction in manual editing time. Q-INT: Accelerated application readiness for production by contributing to 60+ tasks per month, including feature implementation, bug resolution, and enhancements, one month before the expected release.

Summary

Facilitated RLHF model training for Meta, Google, NVIDIA, and other projects, achieving a 25% drop in follow-up queries and a 40% rise in user satisfaction across domains like Python, SQL, and logic. Achieved a 95% accuracy rate in grading model responses and a 98% accuracy rate in task reviews.

Summary

Managed a team of 15 developers across projects, ensuring timely delivery and high-quality output. Conducted orientation and onboarding sessions for new hires, reducing time to productivity from 21 days to 4-5 days. Automated daily task assignments, report generation, and progress reporting, increasing developer productivity to 130+ tasks monthly while saving an average of 12 hours per week in manual updates and scheduling.

Summary

Reviewed and graded over 10,000 model responses with 96% accuracy, ensuring high-quality data for fine-tuning models. Conducted data quality audits, identifying and resolving inconsistencies that improved response precision by 30%.

Summary

Durianpay Homepage: Reduced load time by 70%, cut FCP by 56%, and increased Click-Through Rate (CTA) by 150% using progressive image loading, high compression, and code splitting. Durianpay Dashboard: Created a dashboard for merchants, ensuring visibility of transactions worth 2 million USD per month. Streamlined backend error handling by generating common error handling base classes, sub-classes, and constants, decreasing system downtime due to unresolved exceptions and development time by 14% and 15% respectively. Durianpay API & Docs: Developed a comprehensive API and documentation site, reducing the development time by 15%.

Summary

Launched a new product - Inspekthor - to estimate damages in vehicles using Machine Learning and perform damage inspection through a single unified interface within 1 quarter. Trained a MaskRCNN model on a custom data set of 68,000 images of cars; achieved more than 98% precision for scratch detection. Ran field tests with more than 20 field agents, evaluated results, and identified scaling challenges.