Rohan Pant
AI/ML Engineer. Building agentic AI systems, LLM-integrated backends, and RAG pipelines.
Summary
Software engineer specializing in agentic AI systems, LLM-integrated backends, and production Node.js/TypeScript services. Builds autonomous agents with Gemini and Claude (tool use, MCP-style integrations, RAG), plus ML automation and AWS-integrated systems. Open to AI Engineer, ML Engineer, and Software Engineering internships in 2027. F-1 OPT eligible.
Experience
Software Developer, Rutgers University, New BrunswickJan 2026 - Present
- Built an agentic AI chatbot from scratch (Node.js, TypeScript, Express, React), integrating LLM APIs with tool use and retrieval to deliver context-aware academic content across departments.
- Designed a modular LLM evaluation framework supporting 4+ models with structured logging, latency tracking, and token-usage metrics for benchmarking RAG quality and instruction-following.
- Implemented RAG to ground responses in structured academic content, enabling automated MCQ, programming, and essay-question generation mapped to learning objectives.
Member of Technical Staff 2, CohesityJul 2022 - Jul 2025
- Engineered a production Auto-Triage system (Python and ML) classifying defects across Product, Automation, and Environment, cutting debugging turnaround by 37%.
- Built backend workflows processing 150k+ log events with NLP-based pattern detection for faster root-cause analysis across distributed systems.
- Owned a full Jenkins CI/CD pipeline and a Grafana monitoring dashboard, reducing manual QA effort by 20%.
Data Science Intern, Alpha AIJun 2021 - Jul 2021
- Built Facial Expression and Speech Emotion Recognition pipelines (Keras, 27k 48x48 grayscale images).
- Used OpenCV for face detection and emotion prediction, reaching ~73% accuracy.
Education
M.S. Data Science, Rutgers UniversitySep 2025 - May 2027, CGPA 4.0
B.Tech, Computer Science & Engineering (Big Data), University of Petroleum & Energy StudiesAug 2018 - May 2022
Projects
Whyzrrepo ↗
A Socratic thinking companion for children that asks one guiding question at a time, enforces tool safety in code, and records each child's learning journey in a private git-backed growth journal.
gitagentClaudeGeminiNode.jsGit
SCAILrepo ↗
An AI-visibility audit and outreach engine that measures whether major assistants recommend a business, identifies the gaps, and turns the analysis into reports, rewrites, and CRM-ready outreach.
PythonOpenAIClaudeGeminiFastAPISupabase
Patchworkrepo ↗
GitHub App that turns a bug report into a targeted regression-test pull request. It reads the repository's existing test patterns, uses Gemini 2.5 Pro to write a test that fails before the fix and passes after it, then creates the branch, commit, and PR. Built solo at the Zero to Agent hackathon.
Gemini 2.5 ProNext.jsOctokitSupabaseVercel
Agenticallyrepo ↗
AI compliance intelligence that lets healthcare staff search accreditation standards in plain English and get precise, cited answers without digging through PDFs.
Claude SonnetMongoDB AtlasVoyage AINext.jsVercel
Rutgers LLM Benchmarkingrepo ↗
Two-phase evaluation platform that benchmarks four local GGUF models across 14 prompts and nine capability categories, then compares them in a PDF RAG pipeline using retrieval, citation-validity, and latency metrics.
TypeScriptllama-cpp-pythonReactExpressGroq
F1 Race Rewindrepo ↗
ML-powered what-if simulator for six historical Formula 1 races. Change a pit-stop lap or tyre compound, then compare the actual race with a lap-by-lap replay driven by predicted lap times.
PythonStreamlitFastF1scikit-learnPandas
Skills
Agentic AI
Claude Agent SDKMCP ServersTool UseGemini tool callingAutonomous workflowsPrompt engineering
RAG & Vector DBs
RAG pipelinesSemantic searchMongoDB AtlasOpenSearchS3 Vectorssentence-transformers
AWS
LambdaBedrockS3EC2
Languages
PythonNode.jsTypeScriptJavaSQL
Backend & DevOps
ExpressREST APIsJenkins CI/CDGitDocker