Robin K Philip

StarXAI Technology LLP

AI/blockchain startup building agentic AI products and crypto-market intelligence tools

• Designed and deployed agentic chatbot solutions for a blockchain-focused AI product, integrating MongoDB for dynamic user data retrieval and personalization.
• Built confidence-scoring models for cryptocurrency news analysis and a CAGR-based crypto-trading backtesting engine, enabling data-driven strategy evaluation.
• Implemented advanced web-search augmentation to ground agent responses in real-time information, improving analysis reliability beyond static training data.
• Streamlined CI/CD deployment pipelines across multiple projects and strengthened DevOps practices, improving release cycles and production system scalability.

Jardyan Inc, Remote

Remote AI company behind AIpublisher.org, an AI-powered academic journal publishing platform

• Led full-stack development and deployment of AIpublisher.org, an AI-powered journal publishing platform built on React.js, Node.js, and PostgreSQL.
• Built an AI agent-powered manuscript screening pipeline automating quality scoring, scope matching, formatting checks, and keyword recommendations – accelerating editorial review cycles.
• Developed end-to-end platform workflows spanning article submission, peer review, publication management, secure payments, and admin controls.

A full-stack Retrieval-Augmented Generation system that lets users upload documents across five domains (technical, research, legal, healthcare, financial) and ask questions answered only from retrieved content — every claim traced back to a source, with domain-specific retrieval and safety strategies tuned to how each document type is actually structured.

• Architected a multi-domain RAG platform (legal, financial, healthcare, research, technical) combining hybrid semantic + lexical retrieval with domain-weighted Reciprocal Rank Fusion and Cohere reranking for citation-grounded, hallucination-resistant answers.
• Built a SQL-RAG pipeline that extracts structured tables from financial filings and generates safety-filtered SQL queries, delivering higher-precision numeric answers than prose-only retrieval.
• Engineered a RAGAS-based evaluation gate wired into CI, automatically blocking merges that regress faithfulness, factual correctness, or context recall – turning LLM output quality into a hard release gate, not a manual check.
• Instrumented full production observability (Prometheus, LangSmith) with token/cost tracking and automated anomaly detection across every query type.

Technologies: FastAPI, React, LangGraph, LlamaIndex, PostgreSQL, Cohere Rerank, RAGAS, LangSmith, Prometheus, Milvus

ApplyAI is a full-stack, multi-agent AI platform that automates the job application process end-to-end. Upload a resume and a job posting, and an 11-agent LLM pipeline analyzes the fit, tailors your resume and cover letter to the role, scores ATS compatibility, and drafts your outreach email — all with a human-in-the-loop review step so you approve every change before it's made.

• Architected an 11-agent LLM orchestration pipeline (Google ADK + LiteLLM) that parses resumes and job postings, performs gap analysis, and auto-generates ATS-optimized, tailored application materials end-to-end.
• Designed anti-hallucination guardrails restricting the resume-tailoring agent to keywords verifiably supported by the candidate’s real experience, eliminating fabricated skills while cutting tailoring time from hours to minutes.
• Built a hybrid deterministic + LLM ATS scoring system, using the deterministic scorer to drive an automated rewrite feedback loop and the LLM scorer for user-facing before/after comparisons.
• Implemented cost-aware tiered model routing (Groq primary, OpenAI fallback) with full LangSmith tracing across 10+ sequential LLM calls per request for production-grade reliability.

Technologies: FastAPI, Next.js, Google ADK, LlamaIndex, Groq/OpenAI, GCP, LangSmith

AuraCV is a full-stack AI platform that turns a resume PDF into a hosted, shareable developer portfolio. It uses an LLM-powered extraction pipeline to convert unstructured resume text into structured, schema-validated data — no manual data entry required.

• AI resume parsing — Upload a resume PDF; an LLM extraction pipeline (LlamaParse + OpenAI Structured Outputs) converts it into structured JSON (work history, education, skills, projects) with zero manual formatting.
• Hallucination-safe extraction — Prompt-engineered to only use information present in the document; unknown fields are left blank instead of guessed or invented
• Instant hosted portfolio — Generates a live, themeable portfolio on a custom subdomain (name.auracv.me) straight from the extracted data.
• Google OAuth + secure data layer — Supabase-backed auth, database, and storage with row-level security.

Technologies: Next.js, Fastapi, Superbase, Llamaindex, OpenAI

Bachelor’s in Computer Application with AI Specialization

Organized and conducted college-level hackathons, managing end-to-end event coordination and participant mentorship. Active participant in the Google Developer Student Program, representing the college chapter and conducting developer sessions. Built hands-on skills in MEAN Stack web development (MongoDB, Express.js, Angular, Node.js) through self-directed projects.