Hey, I’mAyush Chougula.

AI Engineer

I build AI that does useful work — from grounded answers to voice agents that connect with real business workflows. Computer Science at MIT ADT, graduating in 2027.

Explore my work Take a look at my résumé

Pune, India · Building voice AI at Quensulting.

Ayush Chougula, in a suit and glasses, standing in a warmly lit library
Engineer. Curious human.
Open to opportunities
Useful systems. Thoughtful details.A little more about what I do

Engineering focus: Applied AI, Agents & RAG, Voice AI, LLM evaluation, Always learning.

Ideas, made real.

A few things I’ve built. The thinking behind them, too.

All projects

Curious by default.
Thoughtful by design.

I like the space between “what if?” and “it works.” Here’s how I get from one to the other.

Get to know me

  1. Start with the person.

    What are they trying to do? I define the useful outcome, the constraints, and the ways things could go wrong.

  2. Make the system make sense.

    I map the data, service boundaries, model responsibilities, and fallback paths before connecting the pieces.

  3. Build the whole path.

    Interface, APIs, auth, retrieval, and storage. I like making the parts work together as one product.

  4. Try to break it. Then improve it.

    Weak inputs, uncertain answers, latency, and regressions all belong in the evaluation loop.

  5. Ship, listen, repeat.

    Start with a useful version, observe real behavior, and let what I learn shape the next iteration.

Built along the way.

From finding weak points to building systems that speak. Five roles, each adding something to how I work.

2024Present

  1. Association for Cyber Security

    DevSec Intern

    Sep 2024 – Feb 2025Pune, India

    Internship

    Security foundations

    Learn the weak points.

    An early foundation in asking where a system can fail, through hands-on exposure to web and blockchain security.

    • Conducted web application threat modeling to identify attack surfaces and map mitigation strategies.
    • Performed introductory smart contract security assessments for applications with blockchain integration.
    • Carried out web application security testing, building practical exposure across both domains.
    • Threat modeling
    • Web security testing
    • Smart contracts · introductory

    View certificate for Association for Cyber Security (opens in a new tab)

  2. Nexus

    Project Lead — Internship

    Mar 2025 – May 2025Remote

    Internship

    Team leadership

    Build with a team.

    Helping 4–6 engineers move three applications forward together: Nexus Connect, Nexus AI, and Nexus Campus.

    • Led sprint planning, task breakdown, and backlog management across three concurrent production applications.
    • Maintained code review standards and responsible handling of API keys and secrets.
    • Coordinated QA, deployment cycles, and stakeholder updates on timelines, blockers, and scope.
    • Sprint planning
    • Code review
    • Release coordination
    • Quality assurance

  3. Syniris Technologies

    Web Development Intern

    Jun 2025 – Aug 2025Remote

    Internship

    Full-stack delivery

    Own the whole product.

    Taking a notes and deadline platform from interface to deployment, with access control and reliability built in.

    • Developed and deployed the full-stack platform, owning frontend and backend implementation.
    • Implemented Firebase authentication and role-based access control for appropriately scoped user access.
    • Established CI-based code review and production monitoring, and validated APIs with manual and functional Postman testing.
    • Firebase Auth
    • Role-based access
    • Postman
    • CI workflows
    • API testing

    View certificate for Syniris Technologies (opens in a new tab)

  4. Persistent Systems Inc.

    GenAI Intern

    Sep 2025 – Apr 2026Pune, India

    Internship

    Applied GenAI

    Make AI accountable.

    Building AI workflows is one half of the job. Measuring whether their answers hold up is the other.

    • Designed and deployed agentic LLM workflows using embedding-based RAG, structured prompts, and Gemini on Google Cloud.
    • Built evaluation pipelines combining self-reflection scoring, rule-based validators, hallucination detection, and model drift monitoring.
    • Developed Persona Mail to model writing style from email history; refined token usage, latency, and prompt versions using measured feedback.
    • Python
    • RAG
    • Gemini
    • Google Cloud
    • LLM evaluation
    • Prompt engineering

    View certificate for Persistent Systems Inc. (opens in a new tab)Explore Persona Mail

  5. Quensulting AI LLP

    AI Voice Agents & Automation Engineer

    May 2026 – PresentPune, India · Remote

    Current role

    Voice & automation

    Let the systems talk.

    Connecting conversations to action: voice agents, backend services, and business workflows in one end-to-end system.

    • Build and deploy Retell and Vapi voice agents, shaping call flows, intent recognition, and turn-taking; extend voice tooling with Dograh.
    • Connect calls to CRMs, backend services, and follow-up actions through n8n, including handoffs to WhatsApp Business.
    • Integrate ElevenLabs speech synthesis and build FastAPI services for call routing, session state, and persistence.
    • Retell
    • Vapi
    • Dograh
    • ElevenLabs
    • n8n
    • FastAPI
    • WhatsApp Business API

Same curiosity. More to build.

Let’s talk about what’s next

A few proud moments.

Long builds, good teams, and a few wins along the way.

Showing achievements 01 / 04

Swipe or use the previous and next buttons to browse. When this gallery is focused, use the left and right arrow keys, Home for the first achievement, or End for the last.

A little more to the right

Notes from the process.

14 min read · 1 Nov 2026

DeepInfer, Explained: How DeepSeek Serves 155 Trillion Tokens a Day

DeepSeek open-sourced the inference engine behind its own API. What DeepInfer is, how prefill and decode are split, why KV caches live on SSDs for days, how DSpark speculative decoding works, and what the benchmarks show.

11 min read · 24 Oct 2026

Why I Moved My Voice AI Stack to MLX, and Where PyTorch Still Wins

Running speech recognition, text-to-speech, voice cloning and a local language model on one 16 GB MacBook with Apple's MLX: the architecture, the real benchmark numbers, the quantization results, and the sharp edges.

8 min read · 17 Sept 2026

How Semantic Search Actually Works, and Why It Still Misses Error Codes

Embeddings, cosine similarity, hybrid search and reranking, explained with runnable Python. Plus the failure cases that send good questions to the wrong page, and how to measure them.

Read all my blogs

A practical toolkit.

The right tool for the problem, with room to keep learning.

Explore all skills
Currently curious about

Agentic AI workflows, RAG systems, Voice AI agents.

AI/LLM Systems

LangChain / RAG / Gemini / OpenAI API / Claude API / Groq / LLM evaluation

Backend

Node.js / FastAPI / Go / REST APIs / JWT / PostgreSQL

Frontend

React / Next.js / TypeScript / Tailwind CSS / Monaco Editor / Shiki

Voice AI

Retell / Vapi / Dograh / ElevenLabs / n8n / WhatsApp Business API / Qwen3

Good things start
with a conversation.

Open to freelance projects, AI engineering opportunities, and full-time AI / software engineering roles for 2027.

Let’s talk More ways to connect