About
Hi 👋, I'm Ashim, a software engineer focused on Applied AI and Data Science. Over the past 3+ years I've been building ML pipelines, NLP systems, and GenAI applications across the automotive and healthcare industries, mostly in Python, SQL, and Databricks.
This is my corner of the web where I share projects, notes, and things I'm currently exploring. When I'm not coding, you'll find me reading, learning something new, or tinkering on side projects.
Feel free to reach out if you'd like to connect or collaborate on something interesting.
Experience

Software Engineer, Data Platform
Rivian & Volkswagen Group Technologies
May 2025 – Aug 2025
Palo Alto, CA
- Built scalable RAG pipelines on Databricks (Delta Lake, MLflow) using vector embeddings (Pinecone/FAISS) and LangChain to enable intelligent retrieval across large automotive regulatory corpora, reducing manual review effort by 95%.
- Preprocessed and analyzed 50,000+ unstructured customer reviews from Reddit and X (Twitter); built and evaluated an NLP sentiment classifier (85%+ accuracy) and entity recognition pipeline, surfacing 200+ actionable automotive product insights per month and reducing issue response time by 40%.
- Tracked model performance and hyperparameter configurations in MLflow; maintained reproducible experiment logs and version-controlled all pipeline code via Git.
- Built Tableau dashboards to communicate model behavior and insights to senior leadership; partnered with cross-functional stakeholders to accelerate delivery by 20%.

Data Science & AI Teaching Assistant
CGI / University of Louisiana at Lafayette
Jan 2024 – Dec 2025
Lafayette, LA
- Led applied ML and GenAI curriculum with CGI Inc., mentoring 150+ students on statistical modeling, model evaluation, EDA, and LLM fine-tuning; designed hands-on Python labs covering regression, classification, and ensemble methods.
- Delivered comprehensive training on AI frameworks including PyTorch, Transformer architectures, GPT, BERT, and prompt engineering; designed hands-on labs covering machine learning algorithms for 50+ students.

Machine Learning Engineer
FuseMachines, Inc.
Apr 2021 – Nov 2022
New York, NY
- Built and evaluated deep learning models (3D U-Net, PyTorch) for medical image segmentation achieving ~0.85 Dice score; assessed new clinical data sources and preprocessed multi-modal datasets using PySpark for scalable downstream modeling.
- Developed EHR-based disease prognosis pipelines using Cox proportional hazards and XGBoost; improved early-detection AUC from 0.72 to 0.81 and presented results to clinical stakeholders via dashboards.
- Contributed to REST API integrations for model inference; collaborated with IT on deployment workflows, model versioning, and production readiness.
Education

M.S. Computer Science
University of Louisiana at Lafayette
Jan 2024 – Dec 2025
Lafayette, LA
4.0 GPA
Skills
Languages: Python, SQL, C++, TypeScript
AI Frameworks: LangChain, PyTorch, TensorFlow, OpenAI, HuggingFace, LlamaIndex
Machine Learning: Natural Language Processing, Deep Learning, Pre-trained Models, Model Integration
Cloud Technologies: AWS, Azure, GCP, Databricks (Delta Lake, MLflow), Data Lakes
Data & Analytics: Large Datasets, Data Pipelines, Vector Databases (Pinecone, FAISS), ETL
Development: Version Control Systems (Git), Agile Methodologies, CI/CD, Docker, Software Development
Publications
Gaire R.R., Subedi R., Sharma A., Subedi S., Ghimire S.K., Shakya S. (2022) GAN-Based Two-Step Pipeline for Real-World Image Super-Resolution. ICT with Intelligent Applications. Smart Innovation, Systems and Technologies, vol 248. Springer, Singapore.
Certifications
Machine Learning SpecializationStanford University
Data Engineering SpecializationAmazon Web Services (AWS)
Deep Learning SpecializationCoursera
Linear Algebra & Calculus
MIT OpenCourseWare
MCP · Claude Code · Claude API · Agent Skills · Subagents · Amazon Bedrock
Anthropic
Conferences Attended
vLLM Conference & Ray Summit
August 24–26, 2026 · San Francisco, CA

NVIDIA GTC 2026
March 16–19, 2026 · San Jose, CA

Stanford AI+Health 2025
December 9–10, 2025 · Stanford University · Virtual