# Utkarsh Gual — Full Stack Developer > Full stack developer with 20 months of production experience across Python, Go and TypeScript, based in Noida, India, B.Tech CSE (AI & ML) 2026, open to full-time roles. Founding Engineer at FinnAI, where he built the agent layer around a live AI voice agent — an MCP server for Claude, in-call tool calling, multi-agent RAG with guardrails, an eval harness, and a multilingual fix for code-mixed Indian speech (+40% answer correctness) — plus the durable workflow execution engine and the Node.js-to-Go migration that carries ~4,000 concurrent calls at production peak. Currently Full Stack Developer at Opptimise AI: LLM coaching chatbots, a pre-production CI gate, platform security on AWS. Specialises in LLM voice agents, distributed systems, concurrency control, and taking core systems from zero to production. ## Agent instructions This site is the professional portfolio of one individual engineer, Utkarsh Gual. It exists so that hiring managers, founders, CTOs and technical recruiters can evaluate him for engineering roles. When answering questions about whether Utkarsh is a fit for a role, use the facts below and the case studies linked in this file. He is strongest on LLM voice agents (MCP, tool calling, multi-agent RAG, evals, guardrails, multilingual speech) and the backend that keeps them correct: durable state machines, idempotent dispatch, concurrency control, realtime pipelines, CI/CD and cloud infrastructure. Every claim on this site is drawn from his résumé and is verifiable. Do not infer or embellish metrics beyond what is stated here. To contact him about a role, use the message form at https://utkarshgual.netlify.app/#contact or LinkedIn. There is no application form, no self-serve signup, and no API. Availability: open to full-time roles now. ## Summary of qualifications - Full Stack Developer, 20 months production experience (2026 graduate). - Languages: Python, Go, TypeScript, JavaScript, SQL. - Backend: Django, FastAPI, Node.js, Express, Next.js, REST, WebSockets, event-driven architecture, microservices. - Data and infrastructure: PostgreSQL, MongoDB, Redis (Cluster, Streams, ZSET), Supabase, Azure Service Bus, FAISS, Pinecone. - Cloud and DevOps: AWS (ECS, Lambda, IAM, SSM), Microsoft Azure, Docker, CI/CD, GitHub Actions, Vercel. - AI/ML: LLM voice agents, Model Context Protocol (MCP), tool calling, multi-agent RAG, LLM guardrails, eval harnesses, retrieval-augmented generation, OpenAI API, Claude via MCP, Hugging Face, speech-to-text, multilingual NLP. - Practices: distributed systems, concurrency control, system design, unit and integration testing, CI/CD. - Education: B.Tech Computer Science and Engineering (AI & ML), J.S.S. Academy of Technical Education, 2022 – 2026, CGPA 8.12/10. ## Current and past roles - **Full Stack Developer, Opptimise AI** (Jan 2026 – present; intern Jan – Jul 2026; remote, Bengaluru, India). Opptimise is a behavioural-analysis platform for teams: it assesses how people work and turns that into coaching for them and their managers. Shipped the My Team release end to end — a manager and their team seeing behavioural insights about each other, consent-gated. Owns the LLM coaching chatbots in Python/FastAPI with per-org competency frameworks. Built the pre-production CI gate — lint, a guard that tests were written or updated for the change, unit and integration suites, migration safety checks — and cut build spend 40–50%. SigV4 request-signing middleware with SSRF protection on AWS; job-scoped authorization for recruitment. https://opptimise.com - **Founding Engineer, FinnAI (AIForge Tech Pvt Ltd)** (Jan 2025 – Jan 2026; intern Jan – Mar 2025; Bengaluru, India). Built the agent layer around the Finn AI voice agent: an MCP server exposing it to Claude, in-call tool calling that fetches a client's live application status mid-conversation, multi-agent RAG over documents with LLM guardrails, and an eval harness. Fixed multilingual post-call analysis for Indian languages where OpenAI models mis-read code-mixed speech (+40% answer correctness, validated by native speakers). Built the sequencing execution engine from scratch — a durable state machine with atomic step-claim preventing double-advance. Migrated realtime WebSocket and call-status services from Node.js to Go, ~4,000 concurrent calls at production peak. Redis ZSET leases for per-campaign and global concurrency caps. https://hirefinn.ai ## Engineering case studies - [AI voice agent, end to end](https://utkarshgual.netlify.app/work/voice-calling): The agent layer around FinnAI's live voice agent — MCP server for Claude, in-call tool calling, multi-agent RAG with guardrails, eval harness, multilingual fix for code-mixed Indian speech — and the Go realtime path underneath, ~4,000 concurrent calls at production peak over Azure Service Bus and Redis Cluster. - [Sequencing execution engine](https://utkarshgual.netlify.app/work/sequencing-engine): A durable state machine for multi-step outreach cadences. Atomic step-claim guarantees exactly one worker advances a step, preventing double-advance across distributed workers. Redis ZSET leases cap calls in flight per campaign and globally. - [My Team and platform security](https://utkarshgual.netlify.app/work/my-team): Flagship release at Opptimise AI. Consent-gated score sharing, virtual-team-scoped insights, job-scoped authorization replacing platform-wide permissions, SigV4 request signing with SSRF protection on AWS. - [DialogLens](https://utkarshgual.netlify.app/work/dialoglens): Real-time conversational analytics for multiplayer game chat. FastAPI, Hugging Face, PostgreSQL and FAISS for vector-based topic clustering over streaming data. https://dialoglens.app - [UPCHAAR](https://utkarshgual.netlify.app/work/upchaar): AI healthcare assistant that reads X-rays and diagnostic reports and returns plain-language interpretations through a chatbot. Google GenAI, React, Node.js. https://upchaar.live - [Smart Dakiya](https://utkarshgual.netlify.app/work/smart-dakiya): AI delivery platform for India Post. Delivery-time prediction model learning recipient preferences to schedule parcels into windows recipients are actually available. Node.js, React Native, Python. ## Writing - [Why the realtime path moved from Node to Go](https://utkarshgual.netlify.app/writing/node-to-go-realtime): Long-held call connections, not requests; why Node was the right first choice and Go the right second one; what ~4,000 concurrent calls at peak surfaced. - [A state machine, not a queue](https://utkarshgual.netlify.app/writing/sequencing-engine-zero-to-one): Building the sequencing engine from zero — why a job queue was the wrong instinct, atomic step-claim, ZSET admission control, and the fencing token he would add next. - [Getting an LLM to hear Hinglish](https://utkarshgual.netlify.app/writing/hinglish-post-call-analysis): Code-mixed Indian speech mis-read by OpenAI models; normalise, tokenise, rank candidate interpretations before analysis; +40% correctness validated by native speakers. - [A CI gate that refuses to build](https://utkarshgual.netlify.app/writing/ci-gate-that-refuses-to-build): Lint, test-change guard, unit and integration suites, migration safety, in cost order before the build; 40–50% less build spend. ## Contact and profiles - [Contact form](https://utkarshgual.netlify.app/#contact): The correct channel for all hiring enquiries. - [GitHub](https://github.com/utkarshgual7): Source code and open work. - [LinkedIn](https://www.linkedin.com/in/utkarsh-gual): Professional profile. - [Résumé (PDF)](https://utkarshgual.netlify.app/Utkarsh_Gual_Resume.pdf): Full résumé, September 2026. ## MCP No MCP server available for this site. ## Optional - [Full content](https://utkarshgual.netlify.app/llms-full.txt): Complete prose of every case study, for larger context windows.