Open to opportunities

Arijeet Chakravarty

@arijeetchakravarty

Message

Applied-AI architect building reliable LLM agents with evals, cost control, and production discipline.

India

Message

What I'm looking for

I’m looking to build and ship reliable applied-AI/agent systems in production—backed by evals, deterministic-first workflows, and cost/latency guardrails—either as a senior IC or in engineering leadership (Director/Head/fractional) that values trust and reliability.

I’m an engineering leader and hands-on builder with 18 years across fintech, insurance, and regulated SaaS, equally comfortable owning an engineering org or shipping production AI myself. I treat production AI as a workflow, reliability, and trust problem—not just a model problem—and I’m happiest when I can replace brittle extraction with deterministic-first agent behavior.

Right now I solo-ship a GPT-5.1 document-intelligence agent that replaced regex with LLM matching and a “verify, don’t extract” loop: the model judges each finding and confirms or flags it with a reason. I back this with regression and determinism suites, per-call cost tracking, and robust-vs-fast trade-offs to meet real gateway limits. Before consulting, I led engineering at an insurtech—scaling the team 10 → 20+ and lifting delivery predictability 35%—while driving practical production discipline like caching, async fallbacks, and token-cost guardrails. I build my own tooling (commit orchestration, pre-commit hooks, internal CLIs) so agentic coding and quality controls match how the work actually runs.

Experience

Work history, roles, and key accomplishments

Current

Applied-AI Architect

Current

Independent Consultant

Jan 2026 - Present (5 months)

Built a production GPT-5.1 document-intelligence agent for regulated pharma (eCTD) submissions, replacing regex matching with LLM semantic verification. Engineered a “verify, don’t extract” self-check loop with regression evals and per-call cost tracking, and shipped end-to-end integrations (Salesforce, webhooks, auth) on Python/FastAPI and AWS within a 60s gateway limit.

GPT 5.1 Autonomy Agents SDK Prompt & Schema Authoring Python fastAPI Docker

Engineering Lead (Applied AI)

Renegade Insurance

Nov 2021 - Dec 2025 (4 years 1 month)

Led production document-intelligence and VOIP-to-insight pipelines for non-technical users, cutting manual effort by 60%. Scaled engineering from 10 to 20+ and improved delivery predictability by 35%, while migrating EKS to ECS/Fargate to cut monthly AWS spend 77% (~$14K to ~$3.2K).

OCR + LLM Extraction Deterministic Schemas Sentiment And Compliance Processing LLM Reliability Via Caching AWS EKS To ECS Fargate Migration Delivery Predictability

Payments API Consultant

Mastercard

Jun 2018 - Nov 2021 (3 years 5 months)

Served as SME for the Payments API ecosystem across 100+ partner teams, delivering secure API integrations with OAuth and encryption. Built a Java no-code API Developer Utility that cut onboarding by 50%.

Payment Integration OAuth API Security Encryption REST APIs Java Partner Enablement Onboarding

Senior Principal Consultant

Oracle Financial Services Software

Aug 2007 - Jun 2018 (10 years 10 months)

Delivered FlexCube retail and internet banking for tier-1 banks, achieving 95% on-time delivery under strict compliance. Provided deep production systems and integration support for regulated financial services.

FlexCube Retail Banking FlexCube Internet Banking Compliance First Delivery Systems support Integration Release Coordination On Time Delivery

Solo Founder (Arin's World)

Arin's World

Jan 2026 - Present (5 months)

Shipped a live AI-assisted family photo-memory app solo in 2 weeks, using Next.js 15 and Neon Postgres. Benchmarked multiple vision models on a scored 25-image test set and built an eval-driven two-stage captioning pipeline with prompt tuning via self-scoring.

Next.js Neon Postgres Vision Model Benchmarking Ollama Prompt Engineering Two Stage Vision Pipeline Cloudflare R2 Vercel