Manish Shaw

Manish Shaw

AI Deployment Strategist
India
Bengali, Hindi, English

نبذة عني

I am a Class of 2027 undergraduate at the Indian Institute of Technology (IIT) Bombay, pursuing a B.S. in Economics with a Minor in AI & Data Science. My technical expertise spans full-stack engineering, AI/ML, and cloud…

الخبرة

Manager, AI Community

Institute Technical Council, IIT Bombay
حتى الآن

Architected ResumeRater, a high-availability platform for 10,000+ residents using Next.js and FastAPI.
Designed the microservices backend with Dockerized containers orchestrated via Azure Container Apps.
Leading a team of 6+ engineers through rigorous Agile sprints and code reviews.
Established a robust DevOps culture by architecting GitHub Actions pipelines for automated testing and seamless deployments.

AI and Systems Engineer

Longani Consulting · Sharjah, UAE
Apr 2025 - Sep 2025 · 4 أشهر

• RAG Architecture: Architected a production-grade Retrieval-Augmented Generation (RAG) system utilizingLangChain and OpenAI GPT-4; implemented an Ensemble Retriever combining BM25 Keyword Search withCosine Similarity on dense vector embeddings to boost retrieval precision by 35% for complex legal & consulting queries.• Backend & Security: Engineered a high-throughput FastAPI microservice leveraging AsyncIO for non-blockingI/O operations; implemented robust security via JWT-based Role-Based Access Control (RBAC) and optimizeddatabase performance using SQLAlchemy with Pydantic validation and Connection Pooling to sustain concurrency.• Prompt Engineering: Designed sophisticated Chain-of-Thought (CoT) prompt templates and Few-Shot Learningpipelines to extract structured insights from unstructured multimodal data, significantly reducing hallucination rates.• DevOps & Cloud: Migrated infrastructure from GCP to Azure, orchestrating GitHub Actions CI/CD pipeline forautomated container delivery; enabled zero-downtime releases via Azure App Service with traffic-based autoscaling.

AI and Systems Engineer

Longani Consulting, Sharjah, UAE

Architected a production-grade Retrieval-Augmented Generation (RAG) system utilizing LangChain and OpenAI GPT-4.
Implemented an Ensemble Retriever combining BM25 Keyword Search with Cosine Similarity on dense vector embeddings to boost retrieval precision by 35% for complex legal and consulting queries.
Engineered a high-throughput FastAPI microservice leveraging AsyncIO for non-blocking I/O operations.
Implemented robust security via JWT-based Role-Based Access Control (RBAC).
Optimized database performance using SQLAlchemy with Pydantic validation and Connection Pooling to sustain concurrency.
Designed sophisticated Chain-of-Thought (CoT) prompt templates and Few-Shot Learning pipelines to extract structured insights from unstructured multimodal data.
Migrated infrastructure from GCP to Azure.
Orchestrated GitHub Actions CI/CD pipeline for automated container delivery.
Enabled zero-downtime releases via Azure App Service with traffic-based autoscaling.

Realtime Credit-Scoring & Fraud Engine

Inter-IIT Tech Meet 14.0 (National Finals)

Engineered a fault-tolerant streaming pipeline using Pathway and Apache Kafka to process 1M+ daily transactions at

Resume Rater SaaS Platform

AI Community, IIT Bombay

Engineered a production-grade SaaS platform using Next.js with Server-Side Rendering (SSR).
Architected a high-concurrency FastAPI backend utilizing AsyncIO to handle 500+ simultaneous requests.
Designed a robust deployment pipeline using Docker Multi-Stage Builds.
Orchestrated serverless containers on Azure Container Apps with autoscaling to automatically provision replicas based on HTTP traffic spikes.

Bi-Directional ISL Translator

Smart India Hackathon - Govt of India (National Winner)

Architected a custom Seq2Seq neural network leveraging Bi-Directional LSTMs to capture long-term temporal dependencies in complex gesture sequences.
Upgraded the model to Multi-Head Attention.
Engineered a high-performance inference pipeline by converting PyTorch models to ONNX Format and applying Dynamic INT8 Quantization.
Reduced model binary size by 65% to facilitate browser-based deployment.
Achieved real-time throughput of 30 FPS on commodity CPUs.
Eliminated dependency on GPUs.

Intelligent File Retrieval System

Hilti Hackathon (Finalist)

Architected a scalable semantic search engine utilizing FAISS with HNSW Indexing for approximate nearest neighbor retrieval.
Generated dense vector embeddings via KeyBERT and Sentence-Transformers.
Achieved sub-100ms latency on large corpora.
Significantly outperformed lexical Regex matching.

Transaction Network Fraud Detection

iDEA Hackathon - Ministry of Finance (National Finalist)

Constructed tripartite networks with artificial fraud nodes.
Optimized detection using DeepWalk on transaction graphs.
Developed a stacked ensemble model (GNN + XGBoost) and inductive pooling for real-time risk assessment and flagging.

المهارات

Python C++ TypeScript SQL Dart Bash Go FastAPI Flask Next.js React Node.js LangChain LlamaIndex PyTorch TensorFlow Hugging Face RAG GNNs XGBoost MediaPipe OpenCV FAISS Docker Kubernetes AWS Azure GCP Terraform GitHub Actions PostgreSQL Redis Apache Kafka Pinecone Chroma MongoDB Microservices System Design REST gRPC CI/CD Event-Driven Architecture AsyncIO JWT RBAC SQLAlchemy Pydantic Connection Pooling OpenAI GPT-4 BM25
الإبلاغ عن هذا الملف الشخصي؟