Work in progress
Axentis — Agentic Activity Tracker
MERN · LangChain
An agentic desktop assistant that learns user behavior across browsers and terminals to automate workflows, with a memory layer for OS-level tasks and real-time usage insights.
01 — Software Developer
Developer.
I ship across the stack — web, mobile, and server-side — with a bias toward small, sharp, maintainable codebases. Currently building at Paper2Audio, where I work on iOS, Android, web, analytics, performance, and reliability.
Web on desktop · iOS on mobile
Experience
Paper2Audio
Software Developer
Sep 2025 — Present
Remote · Seattle
- Build and optimize web apps with Firebase/Firestore sync — resolving data consistency and contamination issues while improving performance and reliability across clients.
- Instrument analytics with Microsoft Clarity and Google Analytics to evaluate user behavior and optimize UI/UX flows.
- Design and maintain cross-platform iOS and Android apps with a consistent, scalable user experience.
- Own root-cause analysis and ANR incident resolution across all client deployments.
Samsung R&D
PRISM AI Developer & Research Intern
Nov 2024 — Jun 2025
Bengaluru, KA
- Built a full-stack dataset annotation platform — Django backend, React frontend, deployed on a Linux server.
- Integrated XMem++ for multi-view automated image annotation, video object segmentation, and polygon-based detection.
- Added interactive correction with RITM, propagating and refining annotations through memory-based reference frames.
ProCodeBase
Machine Learning Engineer Intern
Dec 2024 — Jun 2025
Remote · United States
- Built a full-stack ReAct multi-agent system with Autogen for large-scale data analysis, featuring a Zep knowledge-graph memory layer and Presidio (Microsoft) data anonymization.
- Developed a partially deterministic agentic workflow for causal inference of marketing data using DiD, FGES, and RDD.
- Ranked top 10% among analysis agents on the BIRD and Spider SQL benchmarks in internal evaluation.
Larsen & Toubro
ML & NLP Intern
Jun — Jul 2024
Chennai, TN
- Standardized construction activity codes in a master database using the Gemini API, Google Cloud, and NLTK.
- Built a semantic search feature over a vector database — cutting manual tagging from 6 hours to 15 minutes.
Intel
Project Intern
Project Internship
Computer Vision · Optimization
- Designed a real-time video analytics pipeline with Intel DLStreamer and OpenVINO for CPU/iGPU setups.
- Optimized the object detection pipeline with Docker and MKL-DNN, analyzing CPU–iGPU latency–throughput trade-offs.
Selected Projects
Work in progress
Federated Recommendation System
PyTorch · Flower · Django · Redis
A full-stack federated learning recommender where clients train locally without centralizing sensitive data — Redis caching and a Django backend.
Work in progress
Ensemble RL Trading System
SAC · TD3 · FinRL
An ensemble RL framework combining SAC, TD3, and portfolio strategies (MCW, GMV, EW) with news sentiment for NIFTY 100 buy/sell signals.
Work in progress
Water & Climate Monitoring Platform
Full Stack · Data Pipelines
A public platform visualizing flooding, rainfall, and climate anomalies, with automated scraping of Indian government portals for real-time updates.
Toolbox
PythonC / C++JavaJavaScriptSQLReactNode.jsExpressMongoDBDjangoPyTorchFirebaseLangChainDockerGitTailwind