Headline

AI/LLM Engineer | Backend Engineer (Python) @ HirePro | LangGraph Agents, RAG, Prompt Engineering, LLM Evaluation | Django, Celery | Speech AI (STT/TTS) | Ex-Samsung R&D

About

I build production LLM systems in Python — agents, retrieval and the AI pipeline behind an automated technical-interview platform running on OpenAI GPT-4o/GPT-5. What that involves: → Agents and retrieval — I am the sole engineer on HirePro's AI Proctoring agent: a LangGraph agent with RAG retrieval (OpenAI embeddings, Redis vector store, header-aware chunking), tool calling and WebSocket streaming, answering candidate queries live during proctored tests. → LLM pipeline — I own core subsystems of an AI interview platform: adaptive question generation, real-time answer evaluation, automated feedback synthesis. Python, Django, Celery, multi-tenant. → Cost engineering — cut LLM operating cost ~40% via task-tiered model routing (GPT-5 nano/mini/full by task complexity), restructuring prompts to qualify for OpenAI prompt caching, and token-level cost telemetry on every LLM, STT and TTS call. → Evaluation before assumption — I built an offline harness that A/B tests candidate models and prompts against real production transcripts. No model migration ships without measured quality evidence. → Reliability — migrated free-text prompts to strict Structured Outputs with Pydantic / JSON Schema contracts, eliminating malformed-model-output failures in production. → Speech + multilingual — interviews across 20+ languages with regional STT/TTS and audio caching, plus an authenticity detection flow that flags AI-assisted and plagiarised answers. Earlier at HirePro I built centralized incident-management APIs unifying 4+ incident types, key rotation for encryption of candidate PII, and enterprise vendor-callback integrations. Before that, at Samsung R&D Institute (PRISM), I developed a spaCy-compatible Hindi language model with FastText vectors over 50,000+ vocabulary terms, powering a Rasa assistant at 93% intent accuracy. Promoted to Senior Software Engineer within 20 months. Open to conversations about AI/LLM engineering and Python backend roles. robinsingh051s@gmail.com

Location

Greater Bengaluru Area

AI stack

claude code

claude

langchain

langgraph

langsmith

openai

Work experience

Senior Software Engineer at HirePro, Jul 2025 - Present · 1 yr 3 mos, Bengaluru, Karnataka, India · Hybrid, • Sole engineer on HirePro's AI Proctoring agent — a LangGraph agent with RAG retrieval (OpenAI embeddings, Redis vector store, header-aware chunking), tool calling and WebSocket streaming, answering candidate queries live during proctored tests. • Own core subsystems of the LLM pipeline (Python / Django / Celery) behind an AI interview platform running end-to-end technical interviews on OpenAI GPT-4o/GPT-5 — adaptive question generation, real-time answer evaluation and automated feedback synthesis. • Cut LLM operating cost by ~40% via task-tiered model routing (GPT-5 nano/mini/full by task complexity), restructuring prompts to qualify for OpenAI prompt caching, and token-level cost telemetry across every LLM, STT and TTS call. • Engineered an offline evaluation harness that A/B tests candidate models and prompts against real production transcripts, gating every migration on measured quality instead of assumption. • Eliminated malformed-model-output failures in production by migrating free-text prompts to strict Structured Outputs with Pydantic / JSON Schema contracts. • Shipped multilingual interviews across 20+ languages with regional STT/TTS and audio caching, plus an authenticity detection flow flagging AI-assisted and plagiarised answers. • Contributing to platform architecture and code review, and mentoring 3+ junior engineers on debugging and system design. Resolved 50+ critical production issues across the assessment and interview flows.

Software Engineer at HirePro, Nov 2023 - Jul 2025 · 1 yr 9 mos, Bengaluru, Karnataka, India · Hybrid, • Built centralized incident-management APIs in Python/Django unifying 4+ incident types across ATS and assessment operations, plus the backend APIs powering the AI Proctoring feature. • Implemented key rotation for encryption/decryption of candidate PII, and a centralized mediaService preventing unauthorized media access across assessment applications. • Delivered enterprise integrations — vendor callback systems and client-specific assessment configurations for a major enterprise client — plus React.js UI work and browser-extension security tooling that hardened assessment integrity.

Software Engineer Intern at Synthesis Systems Inc., Mar 2023 - May 2023 · 3 mos, Bengaluru, Karnataka, India, - Developed authentication, database management, and RESTful CRUD APIs for inventory management of 1,000+ products. - Implemented CSV export functionality and conducted 100% API test coverage in Postman, ensuring reliability before deployment.

Prism Project Intern at Samsung R&D Institute India, Jul 2021 - Feb 2022 · 8 mos, Bengaluru, Karnataka, India, • Developed a spaCy-compatible Hindi language model with FastText word vectors over 50,000+ vocabulary terms, powering a Rasa conversational assistant at 93% intent accuracy. • Awarded a Samsung PRISM excellence certificate for the project.

Education

Bachelor of Engineering - BE, Computer Science, Cambridge Institute of Technology

Public links

github: https://github.com/singh051

leetcode: https://leetcode.com/robinsingh051s