Aditya Kadam
AI Engineer skilled in LLM Applications, Agentic Systems, and Production RAG
0
Tasks completed
–
Average score
0
Certificates
0
Badges
1
Internships & jobs
About
I am an AI Engineer who ships complete production systems, including retrieval-augmented generation, multi-agent orchestration, and the FastAPI, PostgreSQL, and frontend layers around them. At VSky Solutions, I shipped a grounded SOP RAG assistant and architected a multi-agent Outlook-to-ERP automation pipeline. I choose the right tool for each problem: an LLM agent where judgment is needed, a deterministic engine where reproducibility matters, a trained model where the input is visual.
Skills
Experience
AI Engineer · VSky Solution
Aug 2025 – Aug 2026
Internship · Kolhapur
Worked as an AI Engineer at VSky Solutions, building AI-powered applications, backend services, and workflow automation systems. Contributed to HR Lens, an AI-assisted Outlook add-in that streamlines candidate registration and HR workflows through resume parsing, email classification, candidate summarization, and integrated ATS APIs. Worked across LangGraph, n8n, Google ADK, MCP integrations, Pydantic, and LLM APIs, with a focus on structured outputs, workflow orchestration, and exception handling. Also contributed to BO Universe Explorer, developing asynchronous FastAPI services backed by PostgreSQL and SQLAlchemy, alongside a Vue 3 frontend. Worked on data loading, database integration, API workflows, and deployment support. Additionally, developed SQL-based inventory reporting with site-level filtering and KPI summaries. These projects gave me hands-on experience taking requirements into working software, integrating AI models and external services, debugging cross-component issues, and collaborating on deployment and demonstrations. The outcome was practical experience delivering end-to-end AI and software features that support real business workflows.
PythonFastAPIREST APIsBackend DevelopmentGenerative AILLMsAgentic AIMulti-Agent SystemsRetrieval-Augmented Generation (RAG)Self-Verifying RAGLangChainLangGraphRAG EvaluationPrompt EngineeringLLM EvaluationVector SearchEmbeddingsRerankingPostgreSQLpgvectorQdran
Projects
NailCareAI
TensorFlow, Keras, Flask, VGG-16, OpenCV
Built a nail-disease screening app classifying 17 conditions (~88.6% top-1 accuracy) on a VGG-16 backbone; benchmarked alternative CNN backbones with class-weighted loss for underrepresented classes, evaluated with per-class precision-recall and confusion matrices, and deployed via a Flask endpoint.
Nexus
Ollama (Qwen2.5/Llama-3), FastAPI, Kuzu, PostgreSQL, Vue 3/Quasar
Built a persistent multi-agent system simulating a software engineering team producing real code against a real repo; designed a calibration engine fusing verbalized and behavioral confidence per agent per task type on a local Ollama + Kuzu knowledge-graph stack.
Job Prep Engine
Qwen2.5-3B (LoRA), Ollama, FastAPI (SSE), React + Vite
Fine-tuned Qwen2.5-3B with a LoRA adapter, served via Ollama for low-latency CPU inference behind a FastAPI SSE streaming backend, with an intent router sending queries down distinct reasoning paths. Built the React + Vite frontend with an AbortController-based stop control.
Self-Improving Agentic RAG
Python, LangGraph, FastAPI, React, ChromaDB, BM25, Groq/Llama 3.3
Built hybrid retrieval fusing dense and sparse search via Reciprocal Rank Fusion, with cross-encoder reranking. Designed a closed self-improvement loop: feedback on incorrect answers is classified by failure type and used to adjust retrieval strategy and prompting; added a post-generation verification stage.
Education
D.Y. Patil Agriculture and Technical University
B.Tech, Computer Science & Engineering
2022 – 2026 · 9.5/10.0