Headline

Software Engineer | LLMs · RAG · LangGraph · FastAPI | Open to Work — Immediate Joiner

About

Currently available for immediate joining. I build AI systems with production engineering instincts — where latency, caching, and observability are design requirements, not afterthoughts. Recent AI projects: → Omni-Router — Production LLM gateway with semantic caching (FAISS + Redis), intent-based model routing (NearestCentroid classifier benchmarked against 6 algorithms), and async quality evaluation via ARQ + DeepEval. Simple queries hit cache and skip LLM inference entirely. MLflow tracks every latency, cost, and eval metric. → Legal Intelligence — Contract Risk & Dependency Intelligence System built on a LangGraph state machine with hybrid retrieval (Qdrant semantic search + Memgraph graph traversal), multi-hop dependency reasoning across legal corpora, intelligent query routing (semantic / graph / both / direct), self-correcting rewrite loop, real-time SSE execution timeline, and full observability via Arize Phoenix + Ragas evaluation. → Voice AI Assistant — Sub-300ms STT → streaming LLM → TTS pipeline over WebSockets. Concurrent multi-session architecture with session isolation. Engineering background: 3 years building real-time systems, distributed caching layers, and concurrent architectures in production — the same instincts that drive semantic caching, latency budgets, and observability in my AI projects. Certified: RAG & Agentic AI Professional · Generative AI Engineering Professional (IBM, 2025) Specializations: LangGraph · LangChain · CrewAI · AutoGen · Transformers · Fine-tuning · Vector DBs Open to AI Engineer / LLM Engineer roles. ajeetkrup401@gmail.com

Location

Noida

AI stack

langchain

langgraph

ollama

openai

Work experience

Software Engineer at InnovationM, Mar 2024 - May 2026 · 2 yrs 3 mos, Noida, Building real-time data systems, voice-integrated interfaces, and scalable backend architectures for enterprise products. • Integrated Speech-to-Text inference into internal HR platforms — reduced query resolution time by 40% across 200+ daily production requests • Architected a WebSocket-based real-time engine with concurrent session management — reduced pipeline latency by 40%, supporting 150+ daily streams with zero dropped events • Maintained strict latency budgets and performance observability across high-throughput architectures using New Relic

Full Stack Developer at Klimb.io, May 2023 - Mar 2024 · 11 mos, Remote, Built and scaled backend systems and data infrastructure for a high-traffic B2B SaaS platform. • Designed a distributed Redis caching layer — reduced API response times from 800ms to 250ms (68% improvement) • Scaled backend services to 500+ concurrent connections with 99.9% availability through 3x traffic spikes • Engineered analytics pipelines processing 10,000+ events/day with virtualized data rendering

Full Stack Engineer at Klimb.io, May 2023 - Mar 2024 · 11 mos, Developed a Redis caching layer for a high-traffic B2B SaaS recruitment platform, cutting average API response time from ∼800ms to ∼250ms (∼68% improvement) on cached endpoints. Executed backend scaling to handle concurrent traffic spikes; built analytics event pipelines processing thousands of daily events with virtualized data rendering. Managed PostgreSQL data models and optimized SQL queries to support reliable analytics and audit trails.

Teaching Assistant at Coding Ninjas, Oct 2022 - Feb 2023 · 5 mos, Remote, Mentored 100+ students in Data Structures, Algorithms, and Web Development. Reviewed code submissions and led doubt-resolution sessions.

Education

Bachelor of Technology - BTech, Computer Science and Engineering, Punjab Technical University, Jalandhar

Class XII, Science(PCMC), DAV Public School , NIT Campus