The teams I've worked with, the problems I've spent time on, and a few things I've built independently.
01 / EXPERIENCE
Where I've worked.
Engineering, machine learning, and research roles.
January 2026 — PresentCURRENT ROLE
Ford Motor Company
Data Scientist
Co-leading agentic AI system design and development for a 0-to-1 vehicle budget planning product under a $50M+ AI initiative; driving research in Google ADK harnesses, search, prompt optimization (GEPA), and LLM evaluation.
Architected and deployed a GCP Cloud Run orchestrator-worker pipeline processing 25K events/250 GB daily in 30 min with sharding, parallelization, concurrency, idempotent BigQuery merges and Docker/Cloud Build.
Migrated 3 production pipelines from AWS S3 to internal APIs (12K logs/120 GB daily), cutting runtime 86% (22h to 3h) with 3-tier failover, circuit breakers, retries and connection pooling; eliminated manual recovery.
PythonGoogle ADKGCPBigQueryDockerLLM Evaluation
May 2025 — Aug 2025
Ford Motor Company
Data Science Intern
Built and deployed an AI-powered ECU log analytics platform on GCP using Docker and Gemini 2.5 Flash & Pro models, enabling natural language driven code generation for regex-based log analysis reducing workflow time by ~90%.
Optimized inference performance through prompt tuning, caching, and parallelized log processing cutting average query processing time from 10s to under 0.1s per log file.
Implemented a multi-agentic system using Hybrid Search (semantic, metadata, keyword) & RAG with Langgraph & Qdrant to enhance log retrieval and analysis, validated via rule-based checks and LLM-as-a-judge achieving ~85% accuracy.
Applied for a patent for the system’s architecture, covering code generation, hybrid search, and real-time ECU log analysis.
PythonGCPDockerLangGraphQdrantStreamlit
Jan 2025 — May 2025
Columbia University Irving Medical Center
Graduate Researcher
Used pose estimation, dimensionality reduction (CEBRA), and unsupervised learning (KMeans, HDBSCAN) to decode behavioral signifiers in mice, identifying 16 stable clusters linking decision-making to brain state covering 80% of frames.
PythonScikit-learnUnsupervised Learning
Feb 2024 — May 2024
56 Secure
Machine Learning Engineer Intern
Led improvement of the YOLOv5s intrusion detection model training on over 500k images, resolving 30+ on-ground bugs and increasing performance over the previous version, aiding in deployment (ML on hardware) across 500+ AI security cameras.
Developed a Slack bot for metrics generation and automated new image data preparation pipeline for training readiness using Python, SQL, AWS S3, and Spark, boosting data science team productivity by 200% weekly.
Collaborated with the product team to build custom evaluation scripts using YOLOv8 by analyzing 150+ customer incidents to refine product strategy and boost customer satisfaction.
PythonSQLPyTorchAWSSparkLinuxONNX
Jan 2023 — Oct 2023
Vellore Institute of Technology
Undergraduate Researcher
Co-authored an IEEE Access Journal paper 'SSBTCNet,' a semi-supervised brain tumor classification model using autoencoder features and fuzzy logic, surpassing leading image classification networks in accuracy, precision, and efficiency.
PythonTensorFlowOpenCV
Jun 2022 — Jul 2022
Cyient
Data Scientist Intern
Structured and optimized ML models using Tesseract OCR, LabelImg, and YOLOv4 in PyTorch to extract geospatial data and detect engineering symbols from PDFs, cutting computation time by 80% and integrating cross-functional feedback.
PythonPyTorchOpenCV
02 / PROJECTS
Things I've built.
Independent experiments, prototypes, and projects I'm still learning from. Explore a project to read more.
03 / EDUCATION
Where I've studied.
A foundation in data science and computer science.