Computer Engineering undergrad @ MHSSCE, Mumbai (May 2026)
Building and stress-testing AI systems end-to-end — from MCP server design to structured capability evaluations.
LinkedIn • adnanbardgujar@gmail.com • Google Developer • Google Cloud Skills Boost — Gold League
- BE Computer Engineering | CGPA: 8.30
- Interned at Aiolos Cloud Solutions (Jul 2025 – Apr 2026) — architected, evaluated, and documented MCP tools for production AI systems
- Focused on: LLM integration, MCP server architecture, prompt engineering, and AI evaluation methodology
- I don't just build AI systems — I break them first, find the failure modes, and document what production actually needs
- 📍 Mumbai, India
LLM Integration → Claude, LLaMA 3.3 70B, prompt engineering, RAG pipelines
MCP Architecture → Server design, tool integration, access controls, safety boundaries
AI Evaluation → Pass/fail frameworks, adversarial testing, capability boundary reports
Backend / Fullstack → Python, Flask, React, Node.js, MySQL, MongoDB
Cloud → GCP (Certified), AWS, OCI (Certified)
Python · React · LLaMA 3.3 70B · MCP
Full-stack research engine integrating 4 AI tools (web search, scraping, PDF extraction, summarisation) on a custom MCP server. Cut manual research time by 60%. Built a capability evaluation framework with adversarial testing, achieving 82% accuracy across multi-domain queries.
Python · Flask · ML Classifiers
Phishing detection combining rule-based heuristics and ML, achieving 87.8% URL accuracy and 99.2% email classification accuracy. Red-teamed with obfuscated URLs and adversarial samples. Built a risk scoring framework for interpretable, human-in-the-loop review.
Python · scikit-learn
Fire risk model using feature engineering and cross-validation, achieving 87% accuracy on held-out test data. Stress-tested across edge-case environmental conditions to map reliability degradation zones.
Let's build something worth building — open to connect, research collaborations, and interesting problems worth breaking.