About
AI Engineer with 3+ years of experience designing and deploying production-scale LLM and AI systems across RAG, multimodal AI, voice AI, recommendation systems, and agentic workflows. Experienced in AI system architecture, LLM orchestration, inference optimization, retrieval, evaluation, and MLOps, using LangGraph, MCP, PyTorch, Kubernetes, and cloud infrastructure. Built systems processing 1M+ images, 10K+ healthcare queries, and 5M+ records/month, with measurable impact including 95% context reduction, 38% lower latency, 70% lower manual effort, and 2× CTR improvement.
Skills & Expertise (81)
Work Experience
AI Engineer
Hub Platform
Nov 2025 - Present
Engineered a Python-based AI SDLC orchestration platform using LangGraph, LiteLLM, MCP, SQLAlchemy, and PostgreSQL, automating 57 workflows across requirements, architecture, development, testing, and release. Built a cross-platform workflow engine supporting 5+ AI development environments and 30 role-specific AI skills, with dependency validation, human approval gates, multi-model routing, and requirements-to-test traceability. Implemented step-level context compression that reduced workflow prompt size by approximately 95%, from 74,862 to 3,544 characters, improving LLM token efficiency, execution cost, and scalability.
Internal BA
Cloumn Mapper
Feb 2025 - Jun 2025
Built an AI-driven Column mapping platform to align columns from 100+ heterogeneous files across multiple financial data sources into organization-defined schemas. Developed a hybrid schema-matching engine using LLMs, semantic retrieval, embeddings, and rule-based validation, achieving 85–92% automated column mapping accuracy while reducing manual mapping effort by 70%.
AI Engineer
JIO
Aug 2023 - Oct 2025
Built a real-time voice AI chatbot with streaming Speech-to-Text + Text-to-Speech, achieving WER: 12.4%. Optimized end-to-end voice pipeline latency from 4.2s to 2.6s (38% faster) using chunk-based processing and Speculative Decoding. Stored structured conversation metadata in PostgreSQL and Redis-based session caching, reducing repeat query latency by 40%. Optimized LLM inference using Continuous Batching, DAF, and Prefix Caching, reducing context consumption by 38%. Created multimodal healthcare assistant using fintuned LLaMA 3.1, OCR, vision model and NER pipelines to process text + medical images, supporting 10K+ clinical queries with 2-5s response latency. Designed a hybrid retrieval system (BM25 + Vector Search) using Elasticsearch + Qdrant, improving response precision by 22%. Applied RAGAS evaluation (Faithfulness: 0.82, Relevance: 0.87) with doctor-in-the-loop validation and safe escalation handling. Orchestrated a temporal forecasting pipeline by separating static and dynamic features, improving wildfire forecast accuracy by 12%. Accelerated model experimentation using distributed training (PyTorch DDP), reducing training time by 45%. Developed a personalized coupon recommendation and ranking system using ALS collaborative filtering and LightGBM, leveraging user, product, and behavioral features to improve Click Through Rate from 2% to 4%. Deployed a complete pipeline with MLflow, Docker, and Airflow for automatic training and tracking of models. Engineered a distributed visual-search system leveraging FAISS indexing search, scaling to 1M+ img maintaining 2–7s retrieval latency. Deployed a scalable FastAPI microservice with autoscaling (Kubernetes), maintaining 99.2% uptime under peak load.
Data Science Intern
Boeing
Jun 2022 - Aug 2022
Constructed a FastAPI real-time analytics pipeline processing 5M+ flight records/month, cutting anomalies/errors by 90%. Optimized PostgreSQL queries with Redis caching, increasing throughput by 70% and cutting latency from 500ms → 90ms. Containerized ML pipelines with Docker and applied prototypes on AWS EC2/S3, reducing setup time by 40%.
Education
B.Tech. in Computer Science - NIT Trichy
2019 - 2023 · Afghanistan
Certifications
No certifications added yet
Interested in this developer?
Profile Score Breakdown
Profile Overview
Skills (81)
Click a skill to find developers with the same skill
