SWE Intern @ OrbitronAI

Kanishka Utagikar

Software Engineering & Applied AI

I design and ship end-to-end applied AI systems, including retrieval architectures and high-performance backend services.

Experience

Software Engineering Intern — OrbitronAI (Bengaluru, India)

May 2026 – Jul 2026

  • Architected a distributed, multi-agentic black-box testing platform leveraging FastAPI, PostgreSQL, and Redis, utilizing Celery to manage parallel asynchronous tasks.
  • Integrated a semantic codebase indexing engine via ClickUp MCP and GitHub MCP to map tickets to repository code, autonomously generating API smoke tests.
  • Implemented an autonomous browser execution agent to bypass 2FA, reducing pre-release bugs by 35% using Gemini 2.5 Pro as a multimodal judge.

AI Intern — MENA AI (Dubai · remote)

Jan 2026 – Apr 2026

  • Decreased average content creation time by 40% by designing an AI-driven conversational pipeline leveraging Temporal for HITL approvals.
  • Reduced API token costs by ~40% by leveraging DSPy to optimize structured prompt execution and mitigate context-bloat.

Education

B.E. Information Technology — Thadomal Shahani Engineering College, Mumbai

Third year · CGPA 8.4/10 · expected 2027

Engineering focus

I build the systems around models: agentic pipelines with LangGraph (routing, reflection, human-in-the-loop), parent-document RAG on Supabase/pgvector, and FastAPI + React stacks with OAuth, streaming, and LangSmith-style tracing—so ideas survive contact with real users and traffic.

Skills

Pulled from my resume—grouped for scanning. Not exhaustive of every side project.

Languages

Python, Java, C++, JavaScript, TypeScript, SQL

Generative AI & NLP

LangGraph, DSPy, LangChain, Hugging Face, Agentic Systems, MCP, LangSmith, Vector DBs

Backend & Orchestration

FastAPI, Pydantic, SQLAlchemy, Alembic, Temporal, WebSockets, OAuth2, Flask

ML & data

PyTorch, TensorFlow, Scikit-learn, Pandas

Databases

PostgreSQL · Redis · pgvector · Supabase · VectorDBs

DevOps, Frontend & Testing

Docker, Docker Compose, CI/CD, Playwright, Selenium, Browser Use, React, Streamlit, Vercel, Git

Achievements

Certifications: Machine Learning A-Z (Udemy) · Neural Networks and Deep Learning (Coursera) · Complete Python Bootcamp (Udemy)

Neural inference demo

Pick a digit to animate a toy network (visualization only, not a trained model).

WAITING FOR INPUT…