Multithreaded Systems, LLM Evaluation and AI Engineer

Amritanshu Yadav Lucknow, Uttar Pradesh

***************@*****.*** LinkedIn GitHub Portfolio Google Scholar +91-775******* EXPERIENCE

Defence Research & Development Organisation (DRDO) — Research Intern Jan 2026 - Jun 2026

• Engineered a native multithreaded Win32 telemetry application for real-time avionics bus monitoring, delivering deterministic execution for mission-critical ground testing.

• Designed synchronized data acquisition pipelines with structured logging, modular concurrency, and hardware-independent simulation for offline validation and debugging.

XelronAI — Software Development Engineer Intern (Remote) Jan 2026 – Mar 2026

• Implemented LLM evaluation pipelines across real-world Python repositories—pull request-level reasoning audits, failure taxonomy classification, and oracle-referenced benchmarking for frontier-model reliability assessment.

• Built Docker-based deterministic evaluation infrastructure with pinned dependencies, CI-integrated pytest validation, and scalable scoring environments for reproducible agent workflow comparison. PROJECTS

RAGOps - Enterprise RAG Platform W Next.js, FastAPI, LangChain, FAISS, PostgreSQL 2026

• Built an enterprise RAG platform with hybrid retrieval (BM25 + vector search), cross-encoder reranking, query routing, citation verification, and 3-provider model orchestration with automatic failover and project-level knowledge isolation.

• Developed AutoRAGBench to benchmark 10 RAG architectures across 35 ground-truth queries, measuring faithfulness, relevance, latency, token cost, caching, and error rates; integrated runtime tracing and retrieval analytics. CyberGuardAI - AI Cybercrime Investigation Platform W LangGraph, Gemini Vision, FastAPI, Next.js 2026

• Orchestrated a 7-agent LangGraph workflow spanning evidence analysis, threat intelligence, recovery, and prevention, with real-time WebSocket streaming and coordinated agent execution in an interactive mission-control UI.

• Integrated Gemini Vision OCR for evidence extraction, React Flow attack-graph visualization, and 3-source OSINT analysis

(DuckDuckGo, Reddit, X) to investigate and cross-reference cybercrime intelligence from multiple sources. AuditAI - LLM Reliability & Observability Platform W FastAPI, Next.js 14, PostgreSQL, Python SDK 2026

• Built a production SaaS platform (“Datadog for LLM Agents”) with a deterministic 5-point engine scoring hallucination, faithfulness, injection, compliance, and tool usage via TF-IDF across 30+ patterns — zero inference cost, fully reproducible.

• Integrated aegis-agent (npm safety middleware) for real-time output blocking; shipped model benchmarking across 3 LLM providers (GPT-4, Claude, Gemini), 4-vector adversarial stress-testing, multi-tenant RBAC, and a Python SDK. TribeBharat - Production E-Commerce Platform W Django, React, PostgreSQL, AWS EC2, Razorpay, Cloudflare Dec 2025

• Co-built and launched a live e-commerce platform serving real customers — Django + React with OTP/JWT auth, Razorpay payments, and a production admin dashboard; transitioned ownership to the founding team post-launch.

• Managed AWS EC2 hosting with Nginx reverse proxy and Cloudflare CDN, ensuring HTTPS security, uptime reliability, and scalable server configuration under live production traffic. RESEARCH PUBLICATIONS

IEEE ICAIIC 2026, Tokyo — ML-Assisted Design of a Tunable Microfluidic SIW Antenna W 2026

• Trained Random Forest, XGBoost, and SVR regression models on simulation-generated datasets, achieving 90% reduction in computational overhead vs traditional simulation workflows. IEEE CCNCPS 2025, Dubai — Cvision: Deep Learning Framework for Cataract Detection via Smartphone W Jun 2025

• CNN ensemble (MobileNetV2, VGG19, DenseNet121) achieving 95%+ accuracy; TensorFlow Lite optimization for low-latency edge deployment on Android healthcare devices.

EDUCATION

International Institute of Information Technology, Naya Raipur 2022 – 2026 B.Tech in Electronics and Communication Engineering CGPA: 7.84 City Montessori School, Lucknow 2021

Class XII: 95.8% Class X: 93%

TECHNICAL SKILLS

Languages: Python, C++, Java, JavaScript, TypeScript AI / GenAI: LangChain, LangGraph, HuggingFace Transformers, LLMs, RAG, FAISS, pgvector, TensorFlow, PyTorch, Sklearn, OpenCV Backend & Cloud: FastAPI, Django, Flask, Node.js, Docker, Redis, PostgreSQL, MongoDB, AWS EC2, Nginx, Cloudflare, CI/CD Frontend: React.js, Next.js, TailwindCSS, Framer Motion, React Flow Developer Tools: Git, GitHub, Linux, Postman, JWT, WebSockets, Razorpay, Jira, Bitbucket, Win32 Coursework: Data Structures & Algorithms, Operating Systems, DBMS, System Design, Machine Learning

Contact this candidate