Anom Chakravorty Dithered
anom_

Tech Lead · AI agents & infrastructure

I build AI agents and take them all the way to production: orchestration, retrieval, durable workflows, guardrails and evals. I lead engineering at Miivo, shipping agents for businesses across the UAE.

  • Open to AI Engineer · FDE roles
  • 7× hackathon winner
  • India · remote
How I ship agents Projects GitHub LinkedIn

Focus

Four tracks that feed each other. Use the arrows, or wait.

01 / 04

Agentic systems

Agents that plan, call tools and finish the job. A model in a loop with MCP tools, retrieval and memory, built so a bad step is caught and retried instead of shipped.

  • Tool calling
  • MCP
  • RAG
  • Multi-agent
02 / 04

Durable workflows

Agents fail halfway through things. I run them as durable workflows: every step checkpointed, retried and resumable, with guardrails and evals in the path.

  • Inngest
  • Workflow SDK
  • Guardrails
  • Evals
03 / 04

Forward-deployed delivery

Start from the customer's process, not the model. Scope it with the people who do the work, prototype in days, then wire the agent into their stack. At Miivo that is AI automation for businesses across the UAE.

  • Discovery
  • Prototyping
  • Integrations
  • Handover
04 / 04

Design systems

The part people actually touch. Tokens and components that keep every screen consistent, plus the chat and streaming UI that makes an agent's work readable.

  • shadcn/ui
  • AI Elements
  • Tailwind CSS
  • Next.js

Work

Five years, from an internship to leading the team.

  1. NOW
    Sep 2024 – present
    Dubai, UAE · remote

    Lead Software Engineer

    Miivo · full-time

    Building AI agents and scalable automations for businesses across the UAE.

    • AI agents
    • LLMs
    • Next.js
  2. Nov 2023 – Aug 2024
    Gurugram · remote

    Full Stack Developer

    BitScale · full-time

    Frontend and full-stack work in Next.js, TypeScript and Go with Postgres. Built the Enrichments Panel search by combining semantic and lexical search.

    • Next.js
    • TypeScript
    • Go
    • LLMs
    • Postgres
  3. Aug 2022 – Feb 2024
    Bengaluru · remote

    Full Stack Development Teaching Assistant

    Crio.Do · full-time

    Helped learners debug and solve problems across the MERN stack, and ran live Q&A sessions.

    • MongoDB
    • Express
    • React
    • Node.js
  4. Aug 2021 – Aug 2022
    India · freelance

    Technical Content Writer

    GeeksforGeeks

    Published technical articles and tutorials on web development. Read the articles.

  5. Jan 2022
    San Francisco · remote

    Software Engineer Intern

    Anakin (YC S21)

    Built and maintained Google Data Studio reports, and wrote Python scripts to extract JSONL data from AWS EC2 servers.

    • Python
    • GCP
    • AWS

Day to day

How I use AI at work, as a shell session. Pick a command, or let it run.

anom@agents: ~/day

anom@agents:~$ plan --before-code

  • Write the spec and the failure cases with a model before any code.
  • Ask it to attack the design. Keep what survives.
  • Acceptance criteria become the first eval set.

anom@agents:~$ build --agents 3

  • Coding agents work in parallel branches. I read every diff.
  • MCP servers hand them the repo, the docs and the schema.
  • Harnesses I reach for: Claude Agent SDK and Pi.

anom@agents:~$ eval --every-change

  • No prompt or model change ships without the eval suite.
  • Benchmark quality, latency and cost side by side.
  • Guardrails get tests like any other code.

anom@agents:~$ deploy --durable

  • Long-running agents run as durable workflows: checkpoint, retry, resume.
  • A person approves the steps that can hurt.
  • Traces and token cost are on from day one.

anom@agents:~$ watch --prod

  • Read real traces. Every failure becomes a new eval.
  • Tighten the prompts, tools and guardrails, then redeploy.
  • Same loop the next morning.

Projects

Five things I built and shipped, most of them in a weekend. Each one is live. Tilt the sheets.

Storz

Web3 file storage and sharing. Files are pinned to IPFS and shared by link, with no central server holding them.

GitHub115 stars · 66 forks
Built withJavaScript · IPFS
RecognitionDecentralized Storage Infrastructure Award, Web3 Infinity 2022 (Filecoin)

Lern

A platform to learn anything, with lessons generated by Google's PaLM 2.

RecognitionGrand Prize, Atlas Madness 2023 (Google & MongoDB)

Codz

A coding platform with a GPT-3.5 assistant beside the editor.

Built withJavaScript · GPT-3.5

Hackathons

Prizes from 2021 to 2023, solo and with teams. Tilt a badge to move its light, click to flip it.

Weekend builds · collector's set

  • Atlas MadnessGrand PrizeGoogle & MongoDB · Team Lern
  • Web3 InfinityDecentralized Storage Infrastructure AwardFilecoin · Team Storz
  • IEM HacksGrand PrizeTeam Optimisers
  • Pasckathon 3.0Grand PrizeTeam Optimizers
  • Razorpay FTX2nd Grand PrizeTeam Rixtox
  • Appwrite × Dev.toRunner-upSolo
  • Ureckathon1st runner-upTeam Optimisers
  • GeeksforGeeks GPL3rd PrizeSolo

Agent stack

End to end, in the order it happens: from a customer's problem to an agent running in production.

01 · ScopeWith the customer: map the workflow and agree what done looks like.
  • Discovery
  • Process mapping
  • Success metrics
  • Eval set first
  • Rapid prototypes
02 · OrchestrateThe agent loop and the tools it can call.
  • Vercel AI SDK
  • Claude Agent SDK
  • Google ADK
  • LangChain
  • CrewAI
  • Pi agent harness
  • MCP
  • Tool calling
  • Multi-agent
03 · GroundPut the right context in front of the model.
  • RAG
  • Embeddings
  • Pinecone
  • Vector databases
  • Hybrid search
  • Context engineering
04 · Make durableSurvive crashes, timeouts and half-finished runs.
  • Inngest
  • Workflow SDK
  • Durable workflows
  • Queues & retries
  • Idempotency
  • Human-in-the-loop
05 · GuardLimit what goes in, what comes out and what it may do.
  • Guardrails
  • Structured outputs
  • Prompt-injection defence
  • PII redaction
  • Approval gates
06 · EvaluateMeasure it before and after every change.
  • Evals
  • Testing frameworks
  • Benchmarking
  • LLM-as-judge
  • Regression suites
07 · Deploy & observeShip it, watch it, feed failures back to stage 06.
  • GCP
  • AWS
  • Vercel
  • Docker
  • Serverless
  • CI/CD
  • Observability
  • Tracing
  • Cost & latency budgets

Underneath all of it

Languages & runtime
  • TypeScript
  • JavaScript
  • Python
  • Go
  • Node.js
  • Bun
Frontend & product
  • Next.js
  • React
  • shadcn/ui
  • AI Elements
  • Tailwind CSS
Data & tooling
  • PostgreSQL
  • MongoDB
  • MySQL
  • Git

Let'stalk.

Hiring an AI Engineer or a Forward Deployed Engineer? I am open to both. Email is the fastest way to reach me.

anomchakravorty3008@gmail.com