Pranav Sastry Resume
Pranav Sastry
Bengaluru, Karnataka, India
AI researcher and engineer; incoming Georgia Tech MSCS student, Machine Learning, Fall 2026.
AI researcher and engineer with experience building agentic automation, recommendation prototypes, and machine learning research systems. Co-author of publications in TMLR, AAAI, and IEEE venues. Incoming Master of Science in Computer Science student at Georgia Tech, Machine Learning specialization, Fall 2026.
EDUCATION
Georgia Institute of Technology
Atlanta, Georgia
Incoming Master of Science in Computer Science
Machine Learning
Fall 2026
BMS College of Engineering
Bengaluru, India
Bachelor of Engineering
Computer Science
August 2019–June 2023
EXPERIENCE
Sarvam AI
Bengaluru, India
AI Engineer
January 2026–June 2026
- Built frameworks and pipelines for auto-optimizing agent harnesses and custom code-generation workflows.
- Extended agentic data-collection work toward web automation and browser-operator workflows for human–AI collaboration.
Sarvam AI
Bengaluru, India
AI Engineering Intern
July 2025–January 2026
- Led an agentic scraper-authoring workflow that reduced authoring time from approximately three hours to 15 minutes across four client deployments.
- Designed subagent isolation and execution-based validation; added runtime API detection and deployed with Google Cloud Workflows and Cloud Scheduler.
Lossfunk
Bengaluru, India
AI Resident
April 2025–July 2025
- Founded RecLLM, a conversational product-discovery prototype with item indexing and preference-aware interaction.
- Built an adaptive-memory module intended to retain user preferences and produced a production-ready Shopify-store prototype.
Indian Institute of Science
Bengaluru, India
Research Assistant
September 2023–April 2025
- Conducted research with Prof. Prathosh AP on time-series imputation, data attribution, and reliable machine learning.
- Across seven datasets and four missing-value ratios, ELITS reported 20.97% lower MSE than GPT4TS with 858× fewer parameters in point-masking evaluations.
- Co-authored research on text-guided data attribution and lightweight multivariate time-series imputation.
Indian Institute of Science
Bengaluru, India
Research Intern
January 2023–September 2023
- Co-authored work combining conditional submodular GANs with programmatic weak supervision.
- Built an OpenReview literature-survey tool and an experiment-management workflow for research execution.
CryptoRelief
Bengaluru, India
Python Developer
May 2021–July 2021
- Built APIs, resource-data scrapers, and a Twitter bot for COVID-19 resource tracking.
- Implemented a SHA-256–based search for the tracked-resource workflow, achieving 90% faster retrieval.
PUBLICATIONS
An Efficient Subset Selection Strategy Using Text-Guided Data Attribution to Mitigate Simplicity Bias
2026
Kumar Shubham; Pranav Sastry; Prathosh AP
Transactions on Machine Learning Research
https://openreview.net/forum?id=zZ5YundT95ELITS — Efficient Lightweight Imputation for Time Series
2026
Pranav Sastry; Kalyan Reddy; Sumanta Mukherjee; Vijay Ekambaram; Pankaj Dayama; Prathosh AP
AAAI 2026 Workshop on AI for Time Series Analysis
https://github.com/AI4TS/AI4TS.github.io/blob/main/Camera_ready_AAAI2026/35.elits_camera_ready.pdfFusing Conditional Submodular GAN and Programmatic Weak Supervision
2024
Kumar Shubham; Pranav Sastry; Prathosh AP
AAAI 2024 Main Technical Track
https://ojs.aaai.org/index.php/AAAI/article/view/29423Performance Analysis of Object Detection Algorithms for Waste Segregation
2023
Naman Singh; Pranav Sastry; Niharika B S; Arka Sinha; Umadevi V
2023 Third International Conference on Artificial Intelligence and Smart Energy
https://doi.org/10.1109/ICAIS56108.2023.10073863PROJECTS
AutoPilot
Public prototype for explicit optimization loops in non-differentiable systems.
https://github.com/pranftw/autopilotAiter
Terminal-native AI agent/chat implementation with persistent sessions and MCP integration.
https://github.com/pranftw/aiterNeograd
Educational NumPy deep-learning framework with autograd and gradient checking.
https://github.com/pranftw/neogradODDGen
Synthetic object-detection data pipeline with automatic bounding-box annotations.
https://github.com/pranftw/oddgenSKILLS
Languages Python, TypeScript, JavaScript, C, Bash
Frameworks and tools PyTorch, PyTorch Lightning, Vercel AI SDK, Next.js, Flask, FastAPI, Google Cloud Workflows, Google Cloud Scheduler, Model Context Protocol
Interests Agentic AI, reliable and verifiable software systems, data attribution, efficient deep learning, large language models, retrieval-augmented systems, and API development