Languages

AI/ML & NLP

GenAI

Frameworks & Libraries

Databases, DevOps & Tools

 

An autonomous multi-agent system powered by a LangGraph pipeline, Tavily web search, and Google Gemini LLMs. It researches public companies in real-time and delivers a structured INVEST / PASS / MONITOR verdict with confidence scoring and factor analysis using Server-Sent Events (SSE).

✦ What makes it special:

• 🤖 Multi-Agent Pipeline: Orchestrates complex research tasks using LangGraph.
• 🌐 Live Web Search: Integrates Tavily for real-time data fetching.
• 🧠 Advanced Reasoning: Uses Google Gemini LLMs to output confidence scores and factors.
• ⚡ Real-Time Streaming: Streams verdicts to the frontend instantly via SSE.

 

A Django-based expense-sharing system that automatically works out the minimum number of payments needed to settle group debts using a custom netting algorithm.

✦ What makes it special:

• 🧮 Custom Netting Algorithm: Computes the absolute minimum transactions to settle debts.
• 👥 Group Management: Secure login, friend management, and group creation.
• 📊 Flexible Splitting: Handles exact amounts, percentages, or equal shares seamlessly.
• 🔄 Real-Time Updates: Live balance calculation and optimized Django ORM queries.

 

An autonomous planning agent in Python that follows an Observe → Reason → Plan → Act → Reflect loop to prioritize tasks, catch overload, and reschedule automatically.

✦ What makes it special:

• 🔄 Autonomous Loop: Implements full Observe → Reason → Plan → Act → Reflect cycles.
• 🎯 Rule-Based Scoring: Evaluates task priority and effort without heavy ML overhead.
• 🧠 Memory System: Learns from missed tasks to improve future scheduling estimations.
• ⏱️ Overload Prevention: Automatically reschedules tasks to prevent schedule burnout.

 

An ML-powered prototype that automates candidate screening by converting resumes and job descriptions into vectors to calculate similarity scores. Reached 88% match accuracy and cuts initial screening time by 37%.

✦ What makes it special:

• 📐 Vector Similarity: Converts unstructured resumes and JDs into comparable embeddings.
• ⚡ Efficiency Boost: Demonstrably reduces HR screening time by 37%.
• ⚖️ Fairness Focused: Built with considerations for bias mitigation (gender/experience proxies).
• 🎯 High Accuracy: Achieved an 88% match accuracy on candidate evaluation.