My Resume

Bachelor of Technology in Computer Science and Engineering

Indian Institute of Information Technology, Dharwad

• •Built Argus, a Gemini-powered Google Chat bot for voicebot debugging with tool-calling NL interface, audio latency analysis, and Splunk alert automation that reduced alert noise by 40%; cut debugging from hours to minutes.
• •Hardened LLM execution pipeline using asyncio.wait_for (7s timeout), parallel tool-call guards, and LangGraph greeting-state injection; eliminated silent-call incidents and shipped BlueDart voicebot live with 150,000 monthly calls.
• •Architected nginx WebSocket load balancing achieving 8× capacity (1,024 max connections); built Prometheus/Grafana monitoring with 30s real-time alerts, Ansible roles cutting provisioning from hours to minutes, and LLM timeout metrics.

• •Engineered a scalable RAG-based automation pipeline using n8n, Gemini embeddings, and PostgreSQL (pgvector); parsed email threads, extracted entities, and indexed 384-d vectors, reducing manual processing time by 60%.
• •Integrated LLM summarization with WhatsApp and Telegram alert dispatchers, delivering real-time supply notifications and reports that improved response turnaround time by 50%.
• •Orchestrated end-to-end productionization with automated testing, vector index optimization, and cron-triggered batch ingestion for continuous data freshness across partner supply feeds.

• •Led usability redesign reducing navigation confusion by 40% through revamped Figma workflows, improved information architecture, and design-system components with cross-functional developer handoffs.
• •Conducted user research with 15+ participants, synthesized qualitative feedback into A/B-testable prototypes that increased engagement by 30% and guided product roadmap prioritization.
• •Delivered interactive prototypes and coordinated design documentation enabling the engineering team to ship 3 feature iterations in half the original timeline.

• •Designed and trained a CNN-based plant-disease classifier on 60,000+ augmented images with class-weighting, achieving 90% accuracy and 92% precision; served predictions via Flask REST APIs with sub-200ms inference latency.
• •Optimized backend performance cutting API response time by 50% and server load by 30% through JWT auth, async queues, Redis caching, batched queries, and Docker CI/CD pipelines.
• •Presented the project at the inaugural Infosys Springboard Summit in Bangalore to industry leaders including Nandan Nilekani; received special invite and felicitation for outstanding performance.

• •Constructed a developer cognitive operating system modeling mental workload using 7 research-grade formulas (0.84 correlation to stress), tracking context switches, focus time, and cognitive load across GitHub activity for 51 developers.
• •Deployed on Azure Container Apps with multi-agent intelligence (Focus Agent, Planning Agent, Interrupt Guard); features AI-powered task sequencing, anomaly detection with Z-score severity levels, and cognitive analytics heatmaps.
• •Piloted with 51 developers showing 3–5 hours/week cognitive recovery savings; integrated GPT-4.1 context briefings, adaptive scoring with exponential decay, and personalized outreach emails with time-savings projections.

• •Developed a multi-agent AI platform with 5 specialized agents (Budget Planner, Market Analyst, Tax Advisor, Savings Optimizer, Debt Strategist) using GPT-4.1 and RAG, orchestrated via A2A shared-context protocol with parallel execution.
• •Integrated Fi.Money's live financial API via Model Context Protocol (MCP) with 6 tools for real-time account data, transactions, and investments; achieved 3–8s latency with Socket.IO streaming and LangSmith observability tracing.
• •Shipped on Azure App Service with MongoDB Atlas persistence, dynamic agent scaling (2 agents for simple queries, 5 parallel for complex planning), and session memory ensuring 99.9% uptime across concurrent users.

• •Crafted a MERN-based developer matching platform with swipe-based discovery, vectorized skill embeddings with ANN search, and real-time chat using Socket.IO enabling seamless hackathon team formation and collaboration.
• •Secured the platform with JWT + Google OAuth authentication, integrated Razorpay payment webhooks for premium features, and containerized deployment on Azure via Docker CI/CD pipelines with zero-downtime releases.
• •Delivered a fully responsive mobile-first UI with Tailwind CSS, connection management dashboard, and profile recommendation engine — attracting 200+ beta signups within the first week of launch.

• •Spearheaded development of an AI career assistant using Azure OpenAI GPT-4.1 with RAG (chunking, embeddings, vector store) and dynamic interview question generators tailored to specific job descriptions and skill gaps.
• •Scaled microservices on Azure App Service with queue workers and Azure AI Search, handling 500+ concurrent requests via batching and sharding while maintaining 99% uptime and sub-second response for match scoring.
• •Validated by 100+ users with precision match scoring (0–100%), personalized learning paths, salary insights, and exportable PDF reports — reducing average interview preparation time by 60%.