Agastya Shetty
Ai Engineer
agastya0212@gmail.com+919619455461Mumbai,Indiaurnextdoor.com/users/agastya0212github.com/Auggy212
Summary
AI-native builder who ships LLM-powered software end-to-end — from design through production integration. Hands-on experience integrating LLM APIs into production workflows (NLP contract validation, predictive ABM engine), building Python data and automation pipelines, and using AI coding assistants (Claude, ChatGPT) as a daily part of delivery. Mechatronics engineering foundation with published research in intelligent automation and control systems, currently completing a Master's in AI. Comfortable owning quality of AI-generated output, defining metrics for AI workflows, and working in fast Agile cycles.
Experience 6 mos
- Engaged directly with business problems and workflows to identify high-impact AI use cases, then designed and delivered solutions (NLP contract-intelligence; predictive account-scoring) from concept to production.
- Built LLM and RAG systems hands-on using Python and API integration — sufficient to prototype, demo, and communicate credibly with engineering teams.
- Designed AI systems for production readiness and adoption — grounding, human-in-the-loop, fallback, and evaluation against success criteria.
- Defined KPIs and built dashboards to demonstrate measurable business outcomes and ROI, and presented findings and trade-offs to technical and non-technical stakeholders.
- Owned go-to-market for a new AI service from 0 to 1 — structured the value proposition, built acquisition funnels and outreach, and stood up analytics to prove traction.
- Applied prompt engineering and AI tooling within a fast-moving, high-velocity environment.
Education
Skills
AI & LLM: LLMs · RAG · AI Agent concepts (routing, tool use, human-in-the-loop) · Prompt Engineering · LLM Evaluation · G, Engineering: Python · REST APIs & API Integration · FastAPI, Customer & Commercial: Use-Case Discovery · Workflow Mapping
Projects
built an LLM-driven application that generates questions, evaluates responses, and returns structured feedback
- Engineered an Account-Based Marketing engine that scores and prioritizes high-value B2B accounts using predictive analytics and intent data.
- ▪ Integrated CRM data with AI-generated messaging to automate multi-channel outreach sequences, building the abstraction layer between source data and the LLM.
- ▪ Built real-time dashboards and KPIs to track engagement signals, pipeline progression, and campaign ROI — quantifying the effectiveness of the AI-assisted workflow.
- Built an NLP-based contract validation pipeline using LLMs to extract key clauses, detect anomalies, and flag compliance risks in unstructured legal documents.
- ▪ Integrated LLM API calls into a live Python workflow — handling structured data extraction from raw contract text and reducing manual review time significantly.
- ▪ Owned the quality of AI-generated output, building validation logic to flag missing or non-standard terms and surface low-confidence results for human review.
- ▪ Connected the validation pipeline to existing business workflows, designing the data flow and hand-off points between the application layer and the AI pipeline
Built a retrieval-augmented chatbot over an enterprise knowledge base that classifies incoming queries and routes them across separate retrieval paths for structured (database) and unstructured (document) sources. Implemented a role-based access layer that filters retrieved context against user seniority before generation, ensuring restricted information never enters the prompt. Achieved 85% response accuracy on [N]-question evaluation set. Python, [LLM], [vector DB].
Developed an automated tool to analyze slide decks against quality rules and surface issues programmatically
Languages
- English, Fluent
- Hindi, Native