Hey, I am
Prateek
Aspiring AI Engineer, specialized in Deep Learning models and Agentic AI Development, with a strong foundation in automation and development in SAP ABAP. Pursuing a Master's in Computer Science from University at Buffalo (AI/ML track).
Work Experience
Research Intern
University at Buffalo, SUNY
October 2025 – Present
- Reported 4.44 Mean Opinion Score (MOS) in Speech Audio Synthesis (DiffWave Implementation): Curating the DiffWave pipeline for speech generation using the pre-trained 22.05 kHz model with mel-spectrogram conditioning; inject Gaussian noise via a Markov chain forward process and train reverse diffusion to recover clean audio.
- Vehicle Speed Imputation (Diffusion + GNN/FGTI): Studying and applying the above diffusion process to Graph Neural Networks (GNN) and FGTI for imputing missing values in Buffalo roadside speed-sensor time series; coverage windows: Oct-mid Nov 2022, first half of Dec 2022, Feb 2023; imputation targets: half of Nov 2022, second half of Dec 2022, Jan 2023 (full)
Business Technical Analyst - SAP Development
Deloitte US offices of India
July 2021 – July 2025
- Engineered a PII data-scrambling and compliance automation platform using Python and SAP ABAP, automating manual data-masking workflows and improving enterprise data-privacy compliance across CRM systems by 90%.
- Refactored legacy cloud-based systems in the Custom Code Decommission initiative; decommissioned 35K+ redundant objects, improved system response time by 45%; enabled a $350K project renewal through scalable backend redesign.
- Delivered 110+ production-grade enhancements under Agile and CI/CD pipelines in ServiceOne; strengthened cross-team SLA adherence to 87% through test-driven, maintainable feature development.
- Resolved 70+ ABAP defects using SQL tuning, debugging, and OOP principles, boosting query processing efficiency by 6%
- Recognized with Deloitte's Applause Award for technical excellence and Cross-functional Collaboration on enterprise systems.
ML & AI Intern
Central Tool Room and Training Center (Govt of India)
November 2019 – December 2019
- Trained a real-time human-figure detection model using TensorFlow CNNs; achieved 92% accuracy, reduced image-processing latency by 22%, and lowered false-positive rate by 8% through model tuning.
- Benchmarked classical ML models (KNN, SVM, Naïve Bayes) against deep-learning baselines; achieved 10%+ accuracy improvement through efficient model tuning and dataset preprocessing pipelines using AI/ML frameworks (PyTorch).
Software Development Intern
Tata Motors South Africa
May 2019 – June 2019
- Developed an end-to-end automation system integrating Excel VBA, SQL, and live scanner data, enhancing production-line traceability and shop-floor preprocessing for vehicular manufacturing, increasing production from 11 units to 13 units daily
- Automated the inventory-tally workflow, cutting cycle time by 90% and query latency by 98% across 9 vehicle models.
- Replaced external vendor software with an in-house data management platform, saving ZAR 15,000 annually and improving scalability, maintenance, and operational control.
Core Skills
Interactive 3D View • Drag to Rotate