Soniya Chavan
Chicago,IL
Summary
AI/ML Engineer with 3+ years of experience building and deploying ML and LLM-driven systems across NLP, computer vision, and time-series domains. Improved model performance (F1 upto 0.87) built scalable, real-time pipelines using Python, FastAPI, and AWS.
Overview
4
4
years of professional experience
Work History
AI Engineer
ESOA Technologies
USA
07.2025 - Current
- Led development and deployment of ML/DL models across NLP, time-series, and computer vision, improving model accuracy by 15–20% and reducing manual intervention, resulting in ~20% improvement in operational efficiency
- Built scalable ML pipelines (10K–50K+ records), reducing manual effort by ~30% improving pipeline efficiency and saving 15–20 hours/week, using AI-assisted tools (Cursor, Copilot, Claude) to accelerate development and debugging
- Deployed real-time inference systems using FastAPI and AWS (EC2, S3) with less than 200ms latency, reducing response times and lowering operational overhead and compute costs by ~10–15%
Thesis Researcher
Purdue University
USA
08.2023 - 06.2025
- Built and evaluated LLM-driven pipelines on noisy Twitter datasets (1000+ tweet-day pairs), improving extraction accuracy by ~20% and reducing manual annotation effort by ~40%, enabling scalable and cost-efficient analysis of unstructured health data
- Benchmarked multiple LLMs using precision, recall, and F1-score, achieving up to F1 ~0.87 and recall ~0.91, enabling reliable model selection and consistent evaluation
- Applied prompt optimization and semantic matching techniques to reduce false negatives and improve extraction consistency across large-scale datasets, enhancing downstream analysis quality
Software Developer
24/7 Software
India
08.2022 - 07.2023
- Developed scalable backend systems and data pipelines for real-time incident tracking, improving system responsiveness by 20–25% and reducing latency, leading to faster issue resolution and improved uptime
- Led agile ceremonies (sprint planning, stand-ups) and coordinated cross-functional teams, improving delivery timelines and team efficiency
- Optimized API performance and processed large-scale data using SQL and Python, enhancing system stability, reducing response latency, and improving visibility into key performance metrics
Education
Master of Science - Computer & Information Technology
Purdue University
Hammond, Indiana05.2025
Bachelor of Technology - Information Technology
VIIT, University of Pune
Pune, India05.2023
Skills
- Languages: Python, SQL, C, R ML/AI: Deep Learning, LLMs, NLP, Computer Vision, Time-Series
- Frameworks: PyTorch, TensorFlow, Scikit-learn, Pandas, NumPy Data: ETL, Feature Engineering, Data Processing
- Tools: FastAPI, Streamlit, APIs, Git, Linux, CI/CD, n8n, Workflow Orchestration Cloud/Viz: AWS (S3, EC2), Tableau, Excel
- AI Tools: Cursor, GitHub Copilot, Claude, ChatGPT
Timeline
AI Engineer
ESOA Technologies
07.2025 - Current
Thesis Researcher
Purdue University
08.2023 - 06.2025
Software Developer
24/7 Software
08.2022 - 07.2023
Bachelor of Technology - Information Technology
VIIT, University of Pune
Master of Science - Computer & Information Technology
Purdue University
Publications
- S. Chavan. Detection of Day-Based Health Evidence with Pre-trained LLMs: COVID-19 Symptoms in Social Media Posts. IEEE BIBM 2023, pp. 4208–4212, DOI:10.1109/BIBM58461.2023.10385580
- S. S. Chavan. Recognizing Health Concepts in Twitter Data Using LLMs. Master’s Thesis, Purdue Univ., 2025, DOI:10.25394/PGS.28965833.v1