Rohit Nikam
Available for AI engineering roles

I build LLM systems that hold up under pressure.

AI engineer focused on retrieval pipelines, fine-tuning and autonomous agents. A security research background means I stress-test what I ship before anyone else gets the chance.

View selected work → Resume
Sandip UniversityID · 2027
Rohit Nikam
Rohit Nikam
AI Engineer
AccessALL AREAS
Drag the card
0+LLM systems shipped end-to-end
0BParameter models fine-tuned
0+Retrieval pipelines built
0+Bug bounty programs engaged
About

Engineering intelligence, with a threat model.

Third-year Computer Science student building production LLM systems — and breaking them on purpose, so they don't break in production.

LLM Systems

Fine-tuning large language models with QLoRA, LoRA and DPO. Building production retrieval pipelines on FAISS vector stores, sentence-transformers and instruction-tuned checkpoints.

Models in training

Adversarial Edge

Security is the differentiator. I red-team model outputs and probe for prompt injection, data poisoning and model inversion — hardening systems well before release.

HackerOne active

Engineering

Python-first developer shipping real systems: FastAPI and Flask backends serving inference, JavaScript frontends, Docker-containerised pipelines and Git-managed MLOps workflows.

Shipping weekly

Education

B.Tech Computer Science Engineering at Sandip University, Nashik — expected May 2027. Coursework across machine learning, computer security, data structures and software engineering.

CGPA 8.47 / 10
Toolkit

What I work with.

LLM & AI systems

Large Language ModelsRAG Pipelines QLoRA / LoRADPO Fine-tuning TransformersFAISS Prompt EngineeringQwen2.5 MistralAgentic Workflows

Adversarial & security

Prompt InjectionLLM Red-teaming OWASP Top 10Bug Bounty Research Vulnerability AssessmentCTF

MLOps & infrastructure

Hugging FaceLangChain OllamaDocker FastAPIGitHub Actions Weights & BiasesCUDA

Languages & development

PythonBash JavaScriptFlask Next.jsREST APIs GitLinux
Selected work

Things I have designed and shipped.

Phantom v3.0

Active engagement

An LLM-powered security intelligence framework. A reasoning agent queries a retrieval pipeline indexing MITRE ATT&CK, OWASP and ExploitDB, then selects context-aware strategies across 15+ vulnerability classes autonomously.

  • RAG pipeline built on FAISS and sentence-transformers over three security knowledge bases
  • Chain-of-thought agent for exploit-strategy selection and fingerprint-driven context switching
  • Modular CLI architecture with a fully Git-tracked agentic workflow
PythonLLMsRAGRESTGit
Repository

VulnPrioritizer

Stable

A vulnerability risk scoring engine built on a custom EARS formula — Exploitability, Asset criticality, Risk, Severity — enriched with live NVD/CVE intelligence and ML-informed heuristics.

  • Custom EARS scoring algorithm combining machine-learning signals with CVSS data
  • Live CVE/NVD API integration for automated threat-intelligence enrichment
  • Modular Flask backend with a clean, testable, plugin-extensible REST API
PythonFlaskRESTCVE / NVD
Repository

ThreatMap

Live monitoring

A real-time threat intelligence dashboard that aggregates OSINT feeds, enriches events with LLM-generated summaries and surfaces the signal through a live visualisation layer.

  • LLM-powered event summarisation and severity triage from raw OSINT feeds
  • Modular Python backend with an extensible data-source plugin system
  • REST API with a JavaScript real-time visualisation frontend and configurable alerting
PythonJavaScriptFlaskOSINT
Repository
Experience

Where I have been working.

2026 — PresentIndependent / Open source

AI Engineer — LLM systems & RAG

Building LLM-powered systems end-to-end: data curation, fine-tuning and retrieval pipeline deployment — with adversarial thinking applied at every layer.

  • Fine-tuned Qwen2.5-Coder-7B with QLoRA and DPO on custom adversarial datasets
  • Built FAISS-backed RAG pipelines with sentence-transformers for domain retrieval
  • Developed agentic workflows with chain-of-thought reasoning for autonomous execution
  • LLM red-teaming: prompt injection, jailbreak testing and output-robustness evaluation
2026 — PresentIndependent

Freelance AI & full-stack developer

Delivering AI-integrated solutions for clients — LLM-powered features, REST APIs, JavaScript frontends and Git-managed deployments, with production reliability as the constraint.

2026 — PresentHackerOne · HackTheBox

Security researcher & operator

Authorised researcher on HackerOne and active operator on HackTheBox. Vulnerability research and CTF work that feeds directly back into how I design and defend AI systems.

Let's build something worth breaking.

Open to internships, full-time roles and remote collaboration on AI and security work. I reply within 24 hours.

Get in touch
Emailrohitnikam195@gmail.com GitHubphantombyte06 HackerOneox3sec
LocationNashik, India