Sep 2024 – Jan 2025 · 4 mos · New York, United States · On-site

• Worked under Prof. Darryl Reeves, with a strong emphasis on Bayesian statistics, helping students develop intuition for probabilistic reasoning and inference
• Held weekly office hours to assist 100+ students with Python, probability, statistics, and data analysis concepts
• Provided guidance on homework, coding assignments, and project design, improving student comprehension and outcomes.
• Graded weekly assignments with a focus on fairness and consistency, offering detailed feedback to reinforce learning.

Sep 2024 – Dec 2024 · 3 mos · New York, United States · Hybrid

• Built custom autonomous agents for query resolution with smarter intent classification, contextual reasoning, and a tailored memory module for improved accuracy and continuity.

Apr 2024 – Sep 2025 · 1 yr 5 mos · New York, United States · Hybrid

• Built and deployed a multi-agent system with custom memory modules and educational scaffolding on FastAPI, EC2, and NGINX, enabling real-time adaptive learning for 1,000+ students with a 95% task completion rate.
• Operationalized open-source LLMs (Gemma, Qwen) on AWS SageMaker, reducing inference latency by 25% and improving learning objective alignment by 30%.
• Architected and ran ETL pipelines on HPC using S3 and MongoDB Atlas to process and analyze 14,000+ research images, cutting costs by 40%.

Aug 2021 – Jul 2023 · 1 yr 11 mos · Bangalore Urban, Karnataka, India

• Engineered observability and lineage systems for DataLab, Maersk’s in-house Databricks alternative, enabling $4M in annual savings
• Built a Spark observability suite (monitoring, lineage, config-driven DAGs) as a core component of DataLab
• Designed and optimized a PySpark pipeline to process 155GB of freight data, improving estimation accuracy by 20%
• Developed a full data lineage system in Python with MongoDB and Azure Databricks, cutting failure detection time by 40% and saving 520 engineering hours annually
• Deployed a real-time voice analytics pipeline using OpenAI Whisper for phone transcription, achieving a 30% word error rate and reducing turnaround time for insights

Apr 2021 – Jul 2021 · 3 mos · India

• Built end-to-end traffic violation system using ONNX-optimized fine-tuned YOLOv3 and SSDMobilenetV2 (0.86 mAP); deployed on HPE edge hardware with automated license plate OCR for real-time violation tracking
• Implemented HoloGAN to create a 3D image generator capable of reconstructing 3D facial images from a single 2D photograph with low latency on Azure Cloud.

Jun 2020 – Jul 2020 · 1 mo · Gurugram, Haryana, India

Jun 2019 – Jul 2019 · 1 mo