Aakash Devkar

AI Engineer · Full-Stack Developer · Pune, India

I build AI agents and the full-stack products around them.

Hi, I'm Aakash. AI engineer and full-stack developer. I design LangGraph agents and RAG pipelines, then build the APIs, databases, frontends and deployments that turn them into products. My rule: the LLM interprets, deterministic code decides, and every number is checked before anyone sees it.

0 / 1,790
mismatches between backtest and live decisions (QuantAxis)
689 / 689
backtest trades reproduced by independent SQL (ForexEdge)
17% → 83%
merge-gate block recall with 0 false blocks (Code Review Agent)

Selected work

Projects

Each case study covers the problem, the architecture, the hardest bugs and how the fix was verified.

Agentic AI · Trading

Live

Oct 2026 – Nov 2026

QuantAxis

LLMs interpret, code decides, a human approves: a fail-closed AI trading decision system.

backtest–live decision mismatches
0 / 1,790
backtest–live decision mismatches
LLM eval sign accuracy (42 / 44)
95%
LLM eval sign accuracy (42 / 44)
test coverage on the core packages
98%
test coverage on the core packages
PythonDjangoDjango NinjaCeleryLangGraphLangSmithGroqMCP
Read case study

Multi-agent · RAG

Live

Jul 2026 – Sep 2026

Autonomous Code Review Agent

Four LLM agents review every pull request and block merges on critical bugs.

merge-gate block recall
17% → 83%
merge-gate block recall
false merge blocks
0
false merge blocks
LLM cost per reviewed file
$0.0017
LLM cost per reviewed file
PythonFastAPIAmazon BedrockQdrantGroqLangSmith
Read case study

Microservices · LLM agent

May 2026 – Jul 2026

ForexEdge

A forex paper-trading platform on 19M real minute bars, with a backtester verified to the cent and a research agent whose numbers are machine-checked.

real minute bars
19.2M
real minute bars
trades reproduced in SQL
689 / 689
trades reproduced in SQL
research agent live eval
9 / 10
research agent live eval
NestJSTypeScriptNxKafkaTimescaleDBRedis
Read case study

Full-stack · ML

Live

Oct 2025 – Feb 2026

AgriFlow

Farm-to-buyer crop marketplace with ML price forecasts, online payments and live AWS deployment.

faster p95 on admin orders
21×
faster p95 on admin orders
error rate at 50 users
38.7% → 0%
error rate at 50 users
production p95 on browsing APIs
60 ms
production p95 on browsing APIs
Node.jsExpressPostgreSQLReactTypeScriptExpo
Read case study

Production · Client work

Feb – Jun 2026

Municipal Complaints System

A complaints platform for the Solapur Municipal Corporation, built and deployed end to end during my Apprely internship.

people on the team I led
8
people on the team I led
production deployment
On-prem
production deployment
NestJSTypeScriptTypeORMPostgreSQLReact AdminNext.js
Read case study

Background

Experience

  1. Jun – Aug 2026

    AI Engineer Intern · Genesis Technologies

    • Built a RAG pipeline and multi-agent orchestration with tiered memory for AuraSync, a production AI hiring platform, on Django microservices with Qdrant, ChromaDB, Redis and Kafka.
    • Added a Redis-based idempotency layer so retried agent and pipeline requests never double-process.
    • Hardened a browser-based candidate assessment pipeline (webcam and microphone capture to a facial-analysis service), fixing camera-constraint and audio-signal issues in production.
    • Researched Indian compliance for the product launch (DPDP Act, RPWD Act accessibility) and handled SEO.
  2. Feb – Jun 2026

    Software Developer Intern · Apprely Technologies

    • Led a team of 8 building the Solapur Municipal Corporation complaints system, from API design to production.
    • Built NestJS + TypeORM + PostgreSQL REST APIs with JWT and OTP login, role-based guards and file uploads, plus a React Admin panel for staff.
    • Built the public website with Next.js, Tailwind CSS, shadcn/ui and Material UI.
    • Deployed to the client's on-premises Windows server (Docker on WSL2, Nginx reverse proxy, wildcard SSL) and wrote the deployment runbook.
  3. 2022 – 2026

    B.Tech, Computer Science · NKOCET Solapur (DBATU)

    • Co-inventor on a published patent, Roadwatch: AI-driven monitoring analytics for transportation networks (Application No. 202521048351, June 2025). Built the data pipeline and AI model, designed the architecture and wrote the technical documentation for filing.
    • Major project: AgriFlow, a four-role agri marketplace on web and mobile with ML price and demand forecasts, built and deployed solo.

Toolbox

Skills

AI & agents
LangGraphLangChainMCPRAGEmbeddingsMulti-agent systemsLLM evalsGuardrailsPrompt-injection defencePrompt engineeringLangSmithGroqAmazon BedrockBedrock GuardrailsHugging Face
ML
PyTorchscikit-learnXGBoostFinBERTpandasNumPy
Backend
PythonFastAPIasyncioFlaskDjangoDjango NinjaDjango ChannelsNestJSNode.jsExpressCeleryBullMQKafkaRedis StreamsEvent-driven architecture (outbox)WebSocketsSocket.IOREST APIsJWT/OAuthRBACGitHub AppsPayment gateways (Razorpay)Java + Spring Boot (learning)
Frontend & mobile
ReactNext.jsTypeScriptTailwind CSSshadcn/uiMaterial UITanStack QueryZustandReact Native (Expo)
Data
PostgreSQLTimescaleDBRedisQdrantChromaDBMongoDBSQLiteTypeORMMinIO
Cloud & DevOps
AWS (Cloud Practitioner)S3EC2ECSECS FargateECREFSALBCloudFrontSNSRDSIAMSecrets ManagerSSMCloudWatchTerraformDockerNginxGitHub ActionsBashVercel
Quality & observability
pytestJestVitestHypothesisk6mypyRuffTrivyOpenTelemetryJaegerPrometheusGrafana
Leadership & communication
Team leadership (8 people)Technical writing & runbooksPatent documentationPresenting to clientsHackathon organising

Contact

Open to AI engineering and full-stack roles

Looking for full-time roles building AI agents, backends and full-stack products. Based in Pune; open to remote, relocation, and international teams.

Available to join immediately · IST (UTC+5:30) · open to remote and relocation

aakashdevkar777@gmail.com