AI · Software · Security
Pune, IndiaAn engineer who likes to build.
I build AI applications, cloud infrastructure, and tools for safer systems. Taking ideas from a working prototype to something people can use.
Selected work
A few projects across security, applied machine learning, and generative AI.
contextclip / desktop
concept
Interface conceptA Windows desktop agent that turns copy and paste events into a connected memory of your workflow.
Built an Electron and React workspace over a Python agent, with immutable event history, source-to-destination graph links, local SQLite storage, and optional AI context compression.
A collaborative learning platform that brings cybersecurity lessons, an AI tutor, and proctored assessments into one workspace.
Contributed as part of the GradeU team. The platform connects source-grounded tutoring with adaptive learning paths and learner progress tracking.
An automated malware incident response toolkit for containment, evidence collection, and system recovery.
Unified network, memory, and malware analysis tools with forensic data acquisition, remote execution, and golden image remediation.
An AI/ML building and energy management system for manufacturing facilities, combining forecasting with anomaly detection.
Developing LSTM and XGBoost forecasting models alongside a centralized failure dashboard and automated maintenance alerts.
A text-to-animation application that turns mathematical ideas into explanatory videos using a four-step LLM pipeline.
Built retrieval over 150+ Manim examples, automatic code correction, live progress updates, and a containerized rendering workflow.
Experience
Working across software, infrastructure, and applied AI.
Worked across application development, production deployments, and AWS infrastructure.
Built practical AI pipelines and backend systems for resume analysis, summarization, and events.
Developed perception, mapping, and video transmission systems for autonomous drones.
Built NLP and document automation workflows for insurance processing.
Toolkit
The technologies I use to turn ideas into working systems.
Research & IP
Work across language models, digital forensics, explainable vision, and adaptive learning.
A study of how removing Transformer layers affects six capabilities across four small language model families, and how fine-tuning restores them. We compare recovery strategies while accounting for the training tokens spent planning them.
Once pilot training costs are included, uniform healing matches or outperforms the planned variants across all four model families.
Presented · Publication pending
20266th International Conference on Ubiquitous Computing and Intelligent Information Systems (ICUIS 2026)
An incident response approach that separates remote forensic execution from immutable evidence storage to help prevent contamination during investigation.
Conference listingForthcoming · CRC Press
19 Nov 2026Advances in Intelligent Computing and Communication Technology · AICCT 2025
Combines CNN-based retinal image classification with Grad-CAM visualizations to make model predictions easier to inspect.
Publisher listingUtility patent pending
2026Adaptive learning · Agentic AI
Maps assessment performance to personalized learning paths in real time, connecting learner diagnostics with adaptive content delivery.
Milestones
Hackathons, teamwork, and software put to work beyond the screen.
A little archive of doing things.
Top 8.22% · October edition
Ranked 336 out of 4,090 participants, with a final score of 69.4/100.
View result1st place
Built an AI tool for fake job detection and website verification. Awarded a ₹2,00,000 prize.
1st place · 15 colleges
Developed a multi-agent solution using IBM Watsonx and Cloud, then presented it at IBM Summit in Delhi.
1st place · 20+ teams
Won the inter-departmental machine learning hackathon at Vishwakarma University.
IBM SkillsBuild Impact Showcase · 2025–26
Recognized at the IBM and Edunet Foundation showcase in Bengaluru, presenting our DIP Diet agent with retrieval over a vector store.
Watch DIP Diet demoNavratri 2025 · Pune
Deployed real-time occupancy monitoring at Shri Mahalaxmi Mandir, supporting crowd management for 50,000+ daily visitors.
Continuing education
Continued learning across full-stack development, data science, machine learning, generative AI, and agent workflows.
06 / Get in touch
For an opportunity, a collaboration, or a good technical conversation. I’d love to hear from you.
Say hello