Top picks from my engineering portfolio.

Coldchain Optimization: Logistics Matching Engine

Real-time cold-chain logistics matching platform with geospatial tracking, workflow orchestration, and pricing intelligence across web and mobile experiences.

• Engineered a real-time cold-chain logistics matching engine using a multi-factor scoring algorithm (route overlap, temperature compatibility, capacity, time window, distance deviation) to rank transporter options with confidence scores, enabling optimal shipper-transporter pairing without prior route visibility.
• Designed a distributed request aggregation and workflow system modeling shipper and transporter interactions, supporting multi-parameter matching with eventual consistency across actors and enabling dynamic decision tracking for acceptance/rejection flows.
• Developed an end-to-end tracking and pricing system with geospatial analytics (PostGIS), real-time ETA updates, and cost computation models, delivering interactive shipment visibility across web and mobile platforms using containerized microservices and JWT-secured multi-role access.

CounsConnect | Counsellor-Management Platform

Multi-platform counseling system pairing a Go and Node.js microservices backend with web and mobile clients for real-time, role-based care coordination.

• Architected a multi-platform counseling platform across web and mobile using a microservices architecture (Go + Node.js), enabling seamless cross-platform access and reducing API latency by ~40% through optimized concurrency and event-driven service design.
• Engineered a real-time communication and task management system using Firebase and MongoDB, supporting 500+ concurrent users with sub-100ms message delivery and implementing RBAC-based authorization across multi-role workflows (Admin, Counselor, Patient).
• Designed a secure authentication and data access layer with JWT, bcrypt hashing, and role-based query controls, ensuring high data integrity and protection against common web vulnerabilities while maintaining scalable handling of large user datasets.

MetaQuery | Query Workspace Manager

Multi-tenant Spring Boot backend for managing and evaluating SQL queries across dialects, with isolated per-user workspaces and JWT-secured access.

• Architected a multi-tenant query management backend using Spring Boot, implementing JWT-based stateless authentication and role-based access control across 6+ RESTful endpoints, supporting 500+ daily transactions with secure session handling and high availability.
• Designed a scalable 3-tier architecture (Controller-Service-Repository) with dynamic query lifecycle management, enabling SQL dialect support across multiple databases (PostgreSQL, MySQL, Oracle) and reducing query validation time by ~40% through optimized persistence strategies.
• Engineered a secure workspace isolation model using Spring Security and JPA, enabling multi-user collaboration with controlled access and feedback-driven query evaluation, ensuring consistent data integrity across concurrent sessions.

GraphMind | Financial Memory Assistant

Multi-tenant AI memory assistant with strict user isolation, hybrid retrieval, and production-grade ingestion/reasoning pipelines for financial knowledge workflows.

• Architected a multi-tenant AI memory system with strict user isolation using JWT-based authentication and dual persistence (PostgreSQL + Neo4j), enabling ingestion of conversational data into structured knowledge graphs with automated entity extraction and relationship mapping, ensuring 99.9% data isolation compliance.
• Engineered a hybrid retrieval pipeline combining graph traversal (Neo4j Cypher) and vector search (Milvus ANN), implementing adaptive query strategies and RRF-based ranking to achieve ~45ms average retrieval latency with context-aware multi-signal scoring.
• Developed a production-grade document ingestion and reasoning pipeline supporting multi-format inputs (PDF, images, text) with OCR (PaddleOCR) and LLM-based extraction, enabling fact deduplication, contradiction detection, and citation-aware response generation with audit trails and quality scoring.

Safe Space | Safety Learning Platform

Full-stack, multi-platform safety education system generating age-adaptive AI narratives with emotion-aware feedback for 5,000+ concurrent learners.

• Architected and deployed a full-stack, multi-platform safety learning system supporting 5,000+ concurrent users across web (React) and mobile (React Native), designing 39+ relational models and 100+ RESTful endpoints with JWT/OAuth-based authentication and role-based access control, achieving 99.9% service uptime.
• Engineered an AI-driven content generation pipeline using Google Gemini AI, transforming 50+ real-time news inputs into age-adaptive interactive narratives, while integrating face-api.js-based emotion detection to enable affective feedback and personalized learning experiences.
• Led end-to-end development of a scalable microservices architecture with FastAPI-based data ingestion, real-time caching via Redis, and Dockerized deployment; reduced content generation latency by ~85% and optimized frontend performance to <2s load time through efficient bundling and state management.

InsightGuard | APT Detection Framework

Real-time APT detection framework with attack lifecycle observability, simulation-driven validation, and adaptive anomaly detection.

• Engineered a real-time APT detection framework leveraging ELK Stack and host/network telemetry (Sysmon, Winlogbeat, Packetbeat), enabling end-to-end visibility across the attack lifecycle (reconnaissance to exfiltration) with 19+ production-grade detection rules.
• Designed a multi-stage attack simulation engine replicating APT36 tactics mapped to MITRE ATT&CK, enabling controlled validation of detection pipelines and improving threat coverage across persistence, privilege escalation, and data exfiltration scenarios.
• Developed an ML-driven anomaly detection system using a PPO-based reinforcement learning agent with SHAP explainability, enabling adaptive detection of C2 traffic patterns and reducing false positives in network anomaly detection workflows.

Smart Vehicle Sequencing | Buffer Optimization System

Real-time paint-shop vehicle routing and sequencing system using constraint optimization and discrete-event simulation to minimize color changeovers.

• Engineered a real-time vehicle routing and sequencing system for paint shop operations, optimizing flow across 9 buffer lines from dual input sources using a multi-strategy scoring algorithm to minimize color changeovers and prevent buffer overflow.
• Designed and integrated a constraint optimization pipeline using Google OR-Tools (CP-SAT, MILP), enabling optimal drain-mode sequencing and reducing paint changeover frequency through color-aware scheduling.
• Developed a discrete-event simulation framework using SimPy with configurable arrival rates and dynamic routing rules, enabling realistic plant modeling and improving system throughput via adaptive buffer management and real-time decisioning.

Gen AI Exchange | Artist Marketplace

AI-powered creator marketplace built as a microservices monorepo with multi-role workflows, asynchronous processing, and CDN-backed asset delivery.

• Architected a full-stack AI-powered marketplace using a microservices-based monorepo (Next.js, Express.js, FastAPI), enabling multi-role workflows (Customer, Artist, Investor, Admin) with JWT-based authentication and supporting 50+ relational data models for scalable digital asset commerce.
• Optimized platform performance using Redis caching, Cloudinary CDN, and BullMQ-based asynchronous processing, reducing API latency by ~40% and enabling scalable handling of 1000+ concurrent user requests and background ML-driven workflows.
• Developed end-to-end e-commerce and discovery features including AI-assisted artist profiling, product search, order management, and course delivery, implementing robust validation (Zod, React Hook Form) to achieve 99%+ input accuracy and improved user conversion across responsive interfaces.

PRISM AI | Agentic Collaboration System

Agentic system unifying enterprise event streams into a temporal knowledge graph, using automata-based monitoring to catch workflow violations in real time.

• Architected an agentic AI system that integrates enterprise event streams (Slack, Jira, CI/CD, Confluence) into a unified temporal knowledge graph, enabling context-aware reasoning and eliminating fragmented decision tracking across distributed teams.
• Designed and implemented automata-based monitoring (LTL/Buchi automata) to detect workflow violations and inefficiencies in real time, triggering intelligent agents for lifecycle tracking, duplicate work detection, workload balancing, and knowledge transfer.
• Engineered a hybrid reasoning pipeline using IBM watsonx (LLMs + Orchestrate) to transform raw event signals into explainable insights and automated actions, enabling proactive decision-making, reducing duplicate effort, and improving organizational efficiency.

MetaLens | Metadata Store Viewer

Distributed metadata exploration solution for schema exploration and analytics across 4+ lakehouse formats directly from S3-compatible object stores without external catalog dependencies.

• Designed a distributed metadata exploration solution that allows schema exploration and analytics on 4+ lakehouse formats like Hudi, Iceberg, Delta, Parquet, directly from S3-compatible object stores without any dependency on external catalog systems.
• Orchestrated a microservices-based solution that uses Node.js and FastAPI to provide parallel data services for efficient processing of 100+ GB of data.
• Optimized a solution that uses DuckDB and Apache Arrow to provide efficient metadata querying, resulting in a 60% reduction in scanning data.

Swasthya | Healthcare Intelligence Platform

Full-stack healthcare platform automating multi-role appointment workflows with real-time messaging and an ML-powered medical intelligence layer.

• Architected a full-stack healthcare platform with 7 microservices supporting multi-role workflows (Patient, Lab Head, Lab Assistant, Doctor), automating appointment scheduling and reducing booking time to <2 minutes through streamlined, event-driven workflows.
• Engineered real-time communication infrastructure using Twilio WhatsApp API and asynchronous processing pipelines, enabling bi-directional notifications with sub-second latency and scalable handling of high-volume messaging events.
• Developed an ML-powered medical intelligence layer integrating NLP-based intent classification and OCR-driven report analysis, enabling automated patient query routing and structured data extraction from 100+ lab report formats for actionable analytics.

OCDetect | OCD Assessment Platform

Full-stack ML-powered diagnostic platform combining behavioral assessments and computer vision signals to generate real-time OCD severity insights.

• Engineered a full-stack ML-powered diagnostic platform with 20+ interactive assessment modules, processing 50+ behavioral and questionnaire metrics to generate real-time OCD severity scores (0-100%).
• Developed a real-time computer vision pipeline using MediaPipe and OpenCV via a Flask microservice, capturing 100+ temporal behavioral signals per session for multimodal prediction quality.
• Designed a scalable backend and authentication system with Firebase and MongoDB for persistent user profiles, secure protected routes, session tracking, and insight dashboards.