Featured Projects
AgenticAI-RAG-Chatbot
Problem: Standard LLM chat systems lack domain-grounded knowledge accuracy and multi-step agentic workflow routing.
Solution: Built a production-ready RAG chatbot with LangGraph, FastAPI, Chroma/Pinecone vector databases, and OpenAI/Gemini for answering questions from an Agentic AI eBook.
Built with:
DeltaDocAI
Problem: Manually reviewing and auditing complex engineering design documents for incremental updates is error-prone.
Solution: Developed an AI-powered engineering document comparison platform featuring visual diff generation, grounded RAG chat, delta analysis, OCR, and computer vision.
Built with:
AmosFlowAI
Problem: Sales operations spend excessive manual effort enriching company data, planning personalized outreach, and tracking CRM updates.
Solution: Built an AI GTM automation engine with LangGraph, FastAPI, and Streamlit for automated company intelligence, personalized emails, and CRM sync.
Built with:
InsightForge-AI
Problem: Enterprise research teams require deep, multi-source company insights before high-stakes sales and strategy meetings.
Solution: Engineered a production-grade AI Research Copilot using React, FastAPI, LangGraph, and LangChain for automated company research and meeting prep briefs.
Built with:
Sales Win/Loss Prediction
Problem: Sales pipelines face unpredictable revenue forecasting without predictive lead scoring.
Solution: Engineered an end-to-end Machine Learning pipeline using Random Forest, Scikit-Learn, extensive EDA, feature engineering, and inference on unseen test datasets.
Built with:
AI Job Alert Bot
Problem: High competition for Software Engineering and AI roles requires instant notification when new positions open.
Solution: Built an automated job monitoring system that continuously tracks engineering opportunities across job boards and sends real-time Telegram alerts.
Built with:
NovaTasks
Problem: Modern teams need lightweight, high-performance task managers with secure authentication and persistent state.
Solution: Built a modern full-stack task management application using React, FastAPI, SQLite, and JWT authentication with an intuitive dark-themed UI.
Built with:
ReadTrack
Problem: Readers need seamless article saving directly from iOS Share Sheets with real-time reading progress tracking.
Solution: Built a Read-it-later native iOS app in SwiftUI with WKWebView integration, custom Share Extension, and intelligent reading-time synchronization.
Built with:
instagram-bulk-unlike
Problem: Removing hundreds of historical liked posts on Instagram manually is tedious and time-consuming.
Solution: Created a lightweight browser console script that automates bulk unliking of Instagram posts directly from the browser window securely.
Built with:
PictoAI
Problem: Creators lack fast, secure tools to generate and manage custom high-quality visuals on demand.
Solution: Built a production-grade text-to-image generation platform with multi-provider inference pipelines, prompt caching, Razorpay payment gateway integration, and JWT authentication.
Built with:
Context-Aware Document QA Bot
Problem: Traditional vector search in RAG struggles with exact word matches and lexical queries, leading to incomplete answers.
Solution: Built a production-grade Context-Aware Document Question Answering bot implementing Hybrid Retrieval (combining semantic vector search with BM25 lexical search) and Reciprocal Rank Fusion (RRF) reranking.
Built with:
PlacementGPT
Problem: Students struggle to search and query complex, unstructured placement preparation materials and institutional datasets.
Solution: Created PlacementGPT, an enterprise-grade Retrieval-Augmented Generation (RAG) platform using hybrid retrieval, cross-encoder reranking, and cache-backed queries for high accuracy and speed.
Built with:
AetherGuard Forensic
Problem: Synthetic media and deepfakes threaten digital trust, making verification of video and image content critical.
Solution: Engineered an AI-powered forensic system utilizing EfficientNet and Vision Transformers (ViTs) to detect facial anomalies and digital manipulations across images and video feeds.
Built with:
AforroMart API
Problem: E-commerce inventories suffer from race conditions and sync delays under high-volume ordering workloads.
Solution: Designed a production-ready API for grocery inventory and order management utilizing Django REST Framework, Docker containers, Redis for caching, PostgreSQL, and Celery for asynchronous task processing.
Built with:
MeetMindAI
Problem: Professionals and teams struggle to capture action items and key insights from long, unstructured meetings.
Solution: Developed a TypeScript-powered AI meeting assistant that processes transcripts, extracts actionable items, and generates structured meeting summaries.
Built with:
Persona-Hire-AI
Problem: Recruiters struggle to get rapid, detailed candidate profiles and book interviews seamlessly without back-and-forth emails.
Solution: Developed a candidate representative agent with Hybrid Retrieval (ChromaDB Vector + BM25 Lexical search), RRF reranking (Cross-Encoder), guardrails against jailbreaking, low-latency voice integration (Vapi, ElevenLabs, Deepgram), and autonomous Google Calendar booking.
Built with:
SAP ABAP Employee Management System
Problem: HR systems require strict authorization, database integrity checks, and custom modular reporting outputs.
Solution: Programmed an enterprise Employee Management System using Module Pool programming, SAP Dictionary (DDIC), Object-Oriented ALV Reports, Lock Objects, and BCS email alerts.
Built with:
Project PrivacyHub
Problem: Redesigning cookie and tracking consent boxes is essential to comply with India's DPDP Act (2023) without degrading UX.
Solution: Created an interactive PM case study and mobile prototype of YouTube's data consent experience, aligning user interfaces with strict privacy regulations.
Built with:
CognitiveUX Assignment
Problem: Constructing modular banking portals and custom video player controllers requires highly precise layout matching.
Solution: Built a pixel-perfect replica of the ICICI Bank privileges interface, combined with an interactive YouTube chapter markers system and lead capture forms.
Built with:
SplitEase
Problem: Roommates and travel groups struggle with complex, error-prone manual bill-splitting and balance tracking.
Solution: Built a modern expense-sharing platform that dynamically computes ledger balances on-the-fly, ensuring zero database-level sync anomalies.
Built with:
KnowledgeFlow AI
Problem: Traditional document searching in team task managers relies on exact keywords, missing critical context.
Solution: Created a semantic-search task board leveraging FAISS and SentenceTransformers locally to index and search workspace documentation by meaning.
Built with:
BreatheESG Emissions Platform
Problem: Auditing multi-tenant corporate travel, fuel, and utility logs involves disjointed formats and data gaps.
Solution: Engineered a Django-based enterprise platform parsing mixed-language CSVs, calculating carbon footprints via Haversine distance, and managing reviews.
Built with:
CipherLab-Pro
Problem: Text encryption and decryption utilities often feature clunky, intimidating layouts that discourage non-tech users.
Solution: Developed a sleek, interactive encryption studio featuring a high-fidelity glassmorphic UI, real-time transformations, and dual theme support.
Built with:
AI Visual Storytelling Documentary
Problem: Generating a cohesive AI-generated visual narrative with consistent characters and cinematic quality requires orchestrating disjointed, complex AI models.
Solution: Engineered an end-to-end AI Visual Storytelling pipeline demonstrating character consistency, precise prompt engineering, cinematic scene generation, AI video production, and documentary workflow organization.
Built with:
AI Decision Matrix
Problem: Selecting the right AI model for enterprise systems requires navigating complex cost, latency, and privacy tradeoffs.
Solution: Built an interactive evaluation dashboard and decision framework providing real-time recommendations based on cost analysis and privacy needs.
Built with:
Cloud Kitchen Market Intelligence
Problem: Cloud kitchen operators face high failure rates due to saturated market areas and misaligned regional pricing.
Solution: Developed a market intelligence system using Python to analyze competitor density, customer demand zones, and optimize regional pricing models in Nashik.
Built with:
