# Ajinkya Pahinkar > Software Engineer | AI Infrastructure & Backend Systems | LLM Orchestration, Microservices, PySpark | Azure Location: San Francisco Bay Area, United States Profile: https://flows.cv/ajinkyapahinkar I’m a Software Engineer with a strong foundation in AI, machine learning, and large-scale data systems, currently building production-grade backend and AI-driven platforms at ServiceLink. My work sits at the intersection of AI infrastructure, data engineering, and backend systems: • Designed and built LLM-powered email automation pipelines that detect intent and route high-volume unstructured emails into enterprise workflows, reducing operational cycle times by ~30%. • Migrated complex legacy SQL reporting systems into Microsoft Fabric using optimized PySpark pipelines, implementing Gold-layer Fact/Dim modeling, watermark-based incremental loads, and near-real-time synchronization. • Built scalable PyTest-based API automation frameworks for distributed microservices across mortgage-loan workflows, cutting regression validation time by 20–30%. Previously, I’ve worked on deep learning research across medical imaging, NLP, and remote sensing — training and deploying models using PyTorch (U-Net, Transformers, LSTMs) on large GPU clusters and contributing to Kaggle competitions with 1K+ teams. I’m now looking to work on AI infrastructure, ML platforms, or backend systems that power intelligent products at scale — especially where engineering rigor meets cutting-edge AI. ## Work Experience ### Software Engineer @ ServiceLink Jan 2024 – Present | Pittsburgh, PA • Developed the backend architecture for EXOS Email Automation, creating Python APIs and workflow endpoints. • Migrated legacy reporting to Microsoft Fabric, optimizing 43 SQL-based ETL scripts into PySpark pipelines. • Built a scalable API automation framework for ServiceLink’s EXOS mortgage-loan platform, reducing validation time by 20-30%. ### Teaching Assistant @ Indiana University Bloomington Jan 2023 – Jan 2024 | Bloomington, Indiana, United States Curated assignments and provided mentorship to students from non-technical backgrounds, assisting them in transitioning into the field of data analytics through the course titled 'Intro to Data Analytics ### Computer Vision Research Engineer @ CNS - Indiana University Bloomington Jan 2022 – Jan 2023 | Bloomington, IN • Spearheaded data annotation and augmentation strategies, enhancing blood vessel detection accuracy by 5% in kidney images. • Trained advanced models (U-Net, Res-UNet) on a dataset of 300 kidney images, achieving a Kaggle baseline score with Dice and Jaccard Index scores of 0.86 and 0.87. • Developed a dashboard using Weights and Biases, reducing training time by 30% through optimized model parameter tuning. ### Natural Language Processing Engineer @ Indiana University Bloomington Jan 2021 – Jan 2022 | Bloomington, IN • Developed a web scraper utilizing the Wikipedia API and Python, extracting 400,000 sentences and reducing data collection time by 40%. • Collaborated with Purdue University’s lab to enhance BERT word embeddings extraction, achieving a 20% reduction in output time. • Automated the game Codenames by improving word matching efficiency, obtaining an average cosine similarity score of 0.76. ### Computer Vision Research Engineer @ Indian Institute of Remote Sensing (IIRS), Indian Space Research Organization (ISRO) Jan 2020 – Jan 2021 | Vellore • Enhanced speckle removal in 500 RISAT-1 SAR satellite images using the POAC algorithm in PyTorch, achieving a PSNR of 44 and SSIM of 0.97. • Integrated the POAC model into a user-friendly pipeline, increasing operational efficiency by 15% for geospatial analysis. • Collaborated with ISRO to evaluate processed images, improving validation scores by 7% to meet quality benchmarks. ### Machine Learning Engineer @ Wikilimo Jan 2020 – Jan 2020 | India • Developed a Long Short-Term Memory (LSTM) network in PyTorch to predict pests in cotton and rice crops across 10 Indian regions. • Achieved an 85% accuracy rate and reduced prediction time by 15%, enhancing agricultural decision-making. • Optimized the ML model through quantization, decreasing model size and memory requirements by 4x for mobile deployment. • Utilized the Apriori algorithm to derive weather-pest associations, achieving over 90% confidence in predictive labeling. ### Machine Learning Academic Intern @ National University of Singapore Jan 2019 – Jan 2019 | Singapore Created a Flask web application that accurately classifies tire defects such as tread wear, bulge, sidewall crack, exposed cord and linear air utilizing a CNN model trained in TensorFlow on a dataset of 500 images, with a noteworthy test accuracy of 82%. ### Cloud Intern @ Hewlett Packard Enterprise Jan 2019 – Jan 2019 | Singapore Utilized LSTMs and GRUs to implement a stock price prediction system, using real-time API data stored in Hadoop clusters to facilitate fast and accurate predictions for informed investment decisions, and compared the stocks of Microsoft and Starbucks. ## Education ### Master's degree in Data Science Indiana University Bloomington ### Bachelor of Technology - BTech in Computer Science Vellore Institute of Technology ## Contact & Social - LinkedIn: https://linkedin.com/in/ajinkya-pahinkar - Portfolio: https://ajinkyapahinkar98.wixsite.com/portfolio --- Source: https://flows.cv/ajinkyapahinkar JSON Resume: https://flows.cv/ajinkyapahinkar/resume.json Last updated: 2026-04-11