Hi, I'm Harsh Tripathi — a Machine Learning Engineer and Master's student at the University at Buffalo. I work across applied ML, large-scale data science, and building production-ready applications. My experience spans research, data science, and software development, and I thrive at the intersection of algorithms, systems, and real-world impact. Explore my projects or read my research work to get a deeper insight into my work, ideas, and engineering approach.
Project Highlights
View More
Implemented a protein function prediction model using deep learning techniques. Used ESM2 model as backbone and finetuned it on CAFA 6 dataset.
- AI
- Machine Learning
- Computational Biology
Experience
University at Buffalo
ML Research Engineer
Sep 2024 - Dec 2025
Engineered a custom ray-tracing analytical model using Projection Probability Density Functions (PPDFs) to form system-response matrices for tomographic reconstruction, devised a convex-hull geometric filter to prune redundant calculations and cut computation by ~70% without accuracy loss, re-implemented attenuation integrals using PyTorch vectorization and multi-CPU threading to achieve ~60× speedups, and scaled billion-ray simulations across the UB CCR HPC cluster using MPI and SLURM for distributed processing. Additionally optimized numerical kernels to approach hardware limits of CPU parallelism and memory bandwidth, and co-authored research publications at SNMMI 2025 and IEEE MIC 2025 advancing applied mathematical methods for imaging.
Visit
UCSF Center for Intelligent Imaging
AI Engineer
Jun 2025 - Aug 2025
Developed LLM-driven Agentic AI pipelines for unstructured medical and financial-style reports, creating robust structuring and extraction workflows supported by feedback loops and monitoring systems across 100+ input formats; designed real-time, reproducible inference services using FastAPI and Docker for enterprise-scale deployment; and collaborated closely with research and product teams to ensure ML solutions aligned with domain-specific requirements and operational constraints.
Visit
Piramal Capital
Machine Learning Engineer
Jul 2022 - Aug 2024
Built machine learning pipelines for financial risk modeling and fraud detection, applying ranking algorithms to millions of daily transactions; developed targeting and recommendation models that increased campaign ROI by 18% and strengthened customer acquisition; deployed cloud-native ML services on AWS using Docker and Kubernetes to support real-time, production-grade workloads; and collaborated with quantitative and business teams to translate financial objectives into ML solutions with measurable impact.
visit
Ahmedabad University
AI Researcher - NLP and Computer Vision
May 2022 - Oct 2023
Person Retrieval Using Soft Biometrics leverages NLP and deep learning to identify individuals of interest in surveillance footage based on descriptive textual queries. I developed the end-to-end language pipeline that ingests free-form English descriptions, extracts and standardizes soft-biometric attributes, filters for relevant features, and maps them to the AVSS benchmark schema. The framework is highly scalable, adaptable to varied writing styles, and extendable to broader vocabularies, enabling the vision module to reliably detect matching individuals in video streams. Project completed in collaboration with Dr. Mehul Raval and Dr. Hiren Galiyawala.
Visit
Birla Institute of Technology and Science, Pilani
Teaching Assistant - Statistical Inference and Applications
Aug 2021 - Dec 2021
Teaching Assistant for the Statistical Inference and Applications MATH F353 course.
VisitBirla Institute of Technology and Science, Pilani
Teaching Assistant - Optimization
Aug 2020 - Dec 2020
Teaching Assistant for the course: MATH F212 Optimisation. Helped out in assignment evaluations and doubt solving for a class of roughly 200 students.
Visit
Pass Consulting
Data Science Intern
May 2020 - Jul 2020
Developed a pilot election-forecasting model that links historical constituency-level electoral outcomes with demographic parameters to predict assembly-level results. The system enables PASS Consulting to analyze constituency profiles, quantify demographic influence, and generate early, data-driven forecasts for upcoming assembly elections, supporting strategic planning and decision-making.