Shikshit Gupta
AI Product Developer | Data Scientist
(917) 536-4813
guptashikshit@gmail.com
New York, NY
Analytical, innovative, and motivated data scientist with experience in data analysis, statistical modeling, and machine learning techniques. Proficient in customer data insights, predictive analytics, and data-driven decision-making. Adept at developing strategies for extracting valuable insights from complex datasets and driving data-driven solutions.
Skills & Expertise
Technical Skills
PythonReactNext.jsReact NativeTypeScriptNode.jsNest.jsMySQLPyTorchLangChainLangGraphTransformersQdrantLLM IntegrationAgentic SystemsRAGAWSLambdaBedrockDynamoDBSageMakerMLflowAirflowDockerKubernetes
Core Competencies
LeadershipActive ListeningProblem SolvingPublic SpeakingAdaptabilityVerbal ProficiencyPersuasion
Certifications
Natural Language Processing Specialization
Google Advanced Data Analytics Specialization
Deep Learning Specialization
Professional Experience
- Architected and deployed a production-grade agentic AI platform using LangGraph/LangChain with structured memory layers (working, episodic, semantic) and RAG pipelines to deliver personalized AI-driven decision workflows.
- Designed modular LLM orchestration architecture, integrating backend services (caching, batching, async execution, structured logging, monitoring, retries/fallbacks) to build contextual coaching from uploaded data, achieving ~200ms p50 API latency under peak load.
- Partnered with product managers/customers to translate ambiguous requirements into scalable AI solutions, while establishing governance controls including audit logging, fallback strategies, and structured output validation to ensure reliability.
- Evaluated and benchmarked multiple LLMs and multimodal models across latency, cost, and response quality, iterating on prompts, memory layouts, and retrieval strategies to achieve ~40% lower inference cost and ~2× faster median response times in production.
Skills: LangGraph, LangChain, RAG, LLM Orchestration, Memory Systems, API Design
- Built a real-time Multimodal AI teaching assistant by fine-tuning LLaVA model on teacher audio, slide text, and images, enabling interactive Q&A and detailed slide explanations with text and audio responses to boost engagement for 5000+ students.
- Engineered a highly scalable smart attendance application, featuring integrated GPS tracking, reducing proxy attendance by 90% for 15000+ concurrent users, while maintaining 99.8% uptime and sub-300ms response times for real-time data synchronization.
Skills: LLaVA, Multimodal AI, Fine-tuning, Real-time Systems, GPS Integration
- Developed a financial intelligence platform using RAG and LLMs for premium users, delivering accurate real-time responses about option trading and stock strategies, resulting in 20% faster user onboarding-to-activation conversion.
- Partnered with marketing and UI/UX teams to target financial high-risk customers with personalized outreach, reducing churn by 15%.
- Collaborated with product and SEO teams to optimize ad targeting, driving a 40% increase in organic traffic and 1M+ new user interactions.
Skills: RAG, LLMs, Financial Intelligence, Customer Analytics, SEO Optimization
- Collaborated with cross-functional teams to design and productionize a cloud-native fraud detection microservice on AWS using Python and REST APIs, implementing secure data handling, logging, IAM-based access controls and monitoring to improve detection accuracy by 40%.
- Led end-to-end migration of Disney production workloads from on-prem infrastructure to AWS (EC2, S3, and Auto Scaling) for scalability.
- Architected high-availability and fault-tolerant deployment pipelines, implemented CloudWatch-based monitoring, automated server status reporting workflows to track production uptime, and built failure pattern analytics to enhance operational resilience.
Skills: AWS, Python, REST APIs, Cloud Migration, Microservices, CloudWatch
Projects
Agentic AI Recommendation & Search Platform
Technologies: LangChain, MCP, UCP, Pinecone, Gemini, Google AI Mode
- Integrated Gemini-powered agents to enable real-time best-deals, allowing intent interpretation, cross-source comparison across platforms.
- Built high-throughput ingestion with vector similarity ranking, and long-running agent memory while implementing governance, and auditing.
Real-time AI Fitness Streaming Engine
Technologies: Typescript, AWS, Rate Limiting, Embeddings, OAuth, API Gateway, Prometheus
- Engineered a scalable fitness streaming pipeline using Kafka, supporting 10,000+ concurrent users at <250ms latency for real-time actions.
- Chained LLM access to tools, databases, and external APIs, enabling multi-tool reasoning, autonomous execution, and scalable orchestration.
Education
Master of Science in Artificial Intelligence
Yeshiva University
New York, NY
- Graduated with specialization in AI and Machine Learning
- Focus on advanced AI systems and practical applications
Bachelor of Technology in Information Technology
Lovely Professional University
Punjab, India
- Comprehensive education in Information Technology fundamentals
- Strong foundation in software development and system architecture