Software Engineer - AI Agents - RAG - Data Infrastructure
 |
 |
 |
 |
|---|
MSCS @ Northeastern University. Previously shipped Java REST
APIs and SAP ABAP modules as an intern. Now building AI agents,
RAG pipelines, and data infrastructure, tested the same way:
naive version, guarded version, controlled experiment, real
numbers reported either way.
Stack: Python // TypeScript // Java // SQL // Go // Rust
Mission: catch problems before they become someone else's.
|
SCHEMA-CONTRACT-ENFORCED DATA INGESTION
Dual ingestion pipelines (naive vs. contract-validated), quarantining schema drift before it reaches the warehouse. + 100% catch rate on breaking schema drift
+ 0% false positives
! +15.5ms median validation overhead
- 100% silent corruption (unguarded baseline) |
RAG PIPELINE WITH RETRIEVAL EVALUATION
RAG grounding LLM outputs in retrieved evidence, evaluated against defined benchmarks at every stage, not vibes. + 80%+ citation coverage
+ 70%+ retrieval relevance
+ Evaluated at every pipeline stage |
|
FUNCTIONAL RISK ENGINE, PROPERTY-TESTED
Every risk metric is a pure function, zero shared mutable state, verified with property-based tests across randomly generated inputs. + Pure functions, no shared state
+ Property-based testing (Hypothesis)
+ GARCH(1,1) anomaly detection
+ Built with Claude Code |
MCP-BASED AI COMMERCE AGENT
An AI agent that browses, designs, and completes a full transaction inside ChatGPT using MCP. Built at the InsForge Agentic Dev Tools Hackathon. + Native Stripe checkout
+ Automated tests, every layer
+ 3-person team, Git code review |
| METRIC | UNGUARDED | CONTRACT-GUARD |
|---|---|---|
| Schema drift catch rate | 0% |
🟣 100% |
| False positive rate | n/a |
🟣 0% |
| Silent corruption rate | 100% |
🟣 0% |
| Median overhead | n/a |
🟡 +15.5ms |
[TOP 20] Agent Forge AI Hackathon, Silicon Valley
-> built AutoBrief, AI meeting/project brief generator
[TEAM] Agents You Love 2 Hackathon, Frontier Tower SF
-> built Vibe-Pair, teammate/event-matching agent
[SHIPPED] InsForge Agentic Dev Tools Hackathon
-> built AgentFirst
[TEAM] Frontiers Gen-AI Hackathon, MIT (Breakthrough
Ventures x Google DeepMind)
-> built ManualMind, turning dense industrial
equipment manuals into interactive training:
cited Q&A, learning plans, quizzes, glossary
[BUILT] Caterpillar India Hackathon
-> voice-guided inspection system: step prompts,
dictation, keyword triggers, image capture,
automated report validation
LANGUAGES
AI / ML
DATA & INFRA
WEB
|
MASTER OF SCIENCE, COMPUTER SCIENCE
Northeastern University, Boston, MA
|
BACHELOR OF TECHNOLOGY, INFORMATION TECHNOLOGY
Vellore Institute of Technology, India
|
