Kirtirajsinh Parmar

Artificial Intelligence Intern @ Terracon

About

I am a Computer Science student at the University at Buffalo (GPA: 3.9), passionate about building systems that sit at the intersection of machine learning, high-performance computing, and real-world scientific problems. I contribute to three research labs, each in a different domain which are medical imaging, contactless physiological sensing, and computational optimization. Whether I am writing custom CUDA kernels, designing signal processing pipelines, or applying Bayesian and deep learning methods, I enjoy working close to the stack and pushing what is computationally feasible. Outside of research, I have built and deployed production systems across various experiences, working on real-time data pipelines, anomaly detection, and full-stack analytics tools giving me a strong sense of how research translates into real engineering. I am actively seeking opportunities in machine learning engineering, ML research, systems engineering, and software engineering, and am open to internships and research fellowships.

Country

United States

City

Buffalo

Industry

Computer Software

Skill

High Performance Computing (HPC), Research and Development (R&D), Data Analysis, Program Trading, Automated Trading, Quantification, Quantitative Finance, Statistical Data Analysis, Database & Data Integration, Front-End Development, Automation & Workflows, Financial Reporting, Stochastic Volatility Modeling, Model Validation, Numerical Integration , Fourier-inversion techniques, Convergence and Sensitivity Analysis, Generalized Advantage Estimation, Multiprocessing for parallel environment batching, Convolutional Neural Networks (CNN)

Experience

Artificial Intelligence Intern

Terracon

2026-6 - Present · 4 mos

United States

Research Assistant

The Research Foundation for SUNY

2026-1 - Present · 9 mos

United States

• Built a high-performance optimization framework for a self-collimating SPECT system using Projected Probability Distribution Function (PPDF) ray-tracing to model photon transport. • Applied Bayesian Optimization with Monte Carlo Expected Improvement (MC-EI) and surrogate models (Gaussian Process) to explore high-dimensional detector geometry configurations. • Currently implementing custom CUDA kernels to parallelize PPDF computation, aiming to reduce evaluation latency from minutes to milliseconds to enable large-scale configuration search.

Quantitative Developer

Numerai

2025-6 - 2025-7 · 2 mos

• Engineered a sharded ML training pipeline processing 3.1M encrypted samples weekly with standardized walk-forward validation. • Built automated CI/CD workflows for reproducible model training, artifact versioning, and experiment tracking. • Implemented feature neutralization and stability constraints to improve model robustness in live deployment.

Data Analyst

Pronutec

2024-11 - 2024-12 · 2 mos

Bilbao, Spain

•Developed ETL pipelines for real-time data ingestion from 10,000+ LV gateway devices, ensuring seamless ADMS-SCADA and GIS system integration. • Designed Python/SQL dashboards and Tableau reports for blown-fuse alarms, overload events, and energy-loss analytics, enabling targeted maintenance decisions. • Deployed a Z-score–based anomaly detection algorithm that triggered field-team alerts within 5 seconds of identifying blownfuse alarms, overloads, and energy-loss anomalies, reducing technical and non-technical losses.

Software Development Enginner

Suntech Systems

2024-5 - 2024-6 · 2 mos

Ahmedabad

•Built a modular React application that merges daily sales, cost-of-goods sold, and gross-profit data into an interactive dashboard, supplying finance and operations with instant performance insights. • Scripted server-side CSV and PDF report bundles for weekly P&L, cash-flow trends, inventory aging and backlogs. shrinking preparation time from 4 - 5 h to less than 1 min. • Built a Python/Flask tool with SQL back-end to log and display current stock for more than 100+ SKUs, giving staff one click visibility into inventory. • Implemented an automated low-stock workflow that generates and emails pre-filled purchase orders directly to approved suppliers.