Help & Supportcareers@gravityer.com
Generative AI Engineer with less than a year in RAG Systems & Multi-Agent Orchestration.
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Built and deployed production-grade AI systems, including RAG pipelines, multi-agent microservices, and deep learning models, with full MLOps layers including REST APIs, Docker, and CI/CD pipelines. Experienced in LLM orchestration, transformer architectures, transfer learning, and adversarial robustness. Focused on hallucination-free, deployment-ready applications. Graduating June 2026.
SASTRA University
B.Tech CSE · Cybersecurity & Blockchain Technology
August 1, 2022 – June 1, 2026
OxiqAI
Gen AI Engineer Intern
May 1, 2026 – Present
India
Financial AI Agent: Multi-Agent Investment Microservice
April 1, 2026 – June 1, 2026
• Built a FastAPI microservice with a 5-layer AI pipeline: safety guard, intent classifier, agent router, specialist agents, and SSE streaming across 10 intent types and 3 deployed agents. • Implemented a three-tier LLM fallback: Groq LLAMA 3.3-70B → OpenAI GPT-40-mini → local classifier, ensuring zero downtime on provider failures with ≥85% routing accuracy. • Built three specialist agents: Portfolio Health (CAGR metrics, benchmark comparison), Financial Calculator (DCA, mortgage, FX, deterministic math), and Market Research (live prices, P/E ratio, LLM summary). • Deployed Streamlit UI with portfolio sidebar and live metrics; all 7 pytest tests pass in CI using mocked LLM without any API key.
View ProjectPlacement RAG Assistant
February 1, 2026 – April 1, 2026
Designed and deployed a RAG pipeline over a 500-company dataset with semantic retrieval, grounded LLM responses, and minimized hallucination via strict context-only prompting. • Implemented vector search using ChromaDB with all-MiniLM-L6-v2 ONNX embeddings and LangChain LCEL; retrieves top-10 semantically relevant records per query with full source attribution. • Integrated Groq LLAMA 3.3-70B for inference; wrapped RAG engine in FastAPI with Pydantic validation, health check, latency tracking, and auto-generated Swagger docs. • Containerized with Docker (secrets injection, ChromaDB volume persistence, healthcheck); CI/CD via GitHub Actions with flake8 linting and 6 pytest API tests.
View ProjectCross-Domain Intrusion Detection System (IoT + Cloud)
January 1, 2026 – April 1, 2026
• Achieved 99.52% accuracy on Bot-IoT and 99.86% on CIC-IDS2018 using a Deep Ensemble Neural Network (DENNW) with GAN-based augmentation, NCA feature selection, and Focal Loss for class imbalance. • Designed dual encoder architecture with CORAL domain adaptation: Bot Encoder (15 IoT features → 64-dim) and CIC Encoder (78 cloud features → 64-dim), preserving 100% of features with zero information loss. • Implemented CORAL loss with L2 normalisation, stabilising joint optimisation of CE_bot + CE_cic + 0.01 x CORAL across four models with gradient clipping at norm 1.0. • Validated adversarial robustness under FGSM and PGD attacks; model maintained high detection rates, confirming production-readiness for hostile network environments.
Detection of Cryptographic API Misuse in Source Code
February 1, 2025 – June 1, 2025
• Built an ML pipeline to detect cryptographic API misuses using graph-based embeddings (AST, node2vec, Bag-of-Graphs) for semantic code representation; outperformed Coverity on held-out test sets. • Applied transfer learning via code2vec to improve cross-project generalisation, reducing dependence on project-specific training data. • Used code obfuscation as data augmentation to enhance model robustness against adversarially crafted inputs.
Cultural Fit Analysis
The candidate's projects showcase a diverse range of applications within AI/ML, from RAG assistants and financial agents to intrusion detection and cryptographic API misuse. This breadth of interest and application, combined with an internship in Gen AI, suggests adaptability and a strong drive to explore different facets of AI engineering. The focus on deployment-ready applications and MLOps indicates a practical, results-oriented mindset that would fit well in a fast-paced development environment. However, the academic nature of some projects and the lack of team-based project descriptions (beyond personal projects) mean that direct evidence of team collaboration or leadership is limited.
Soft Skills & Operational Fit
The candidate's project descriptions indicate a strong problem-solving aptitude and a focus on robust, production-ready solutions. The detailed explanations of technical implementations suggest good communication of complex ideas. The internship at OxiqAI further demonstrates an ability to reverse engineer complex systems and propose enhancements, indicating a proactive and analytical approach to work.