Available for opportunities

Deepak
Khimavath

> |

AI Platform & Developer Tools Engineer — I build the infrastructure that helps engineering teams move faster. From agentic AI pipelines to distributed FinTech microservices on Azure.

View My Work Resume Ask AI About Me ↗
98%
Latency Reduced
150+
PRs Reviewed
6
Agent Pipeline
20+
Internal Users
Deepak
Deepak Khimavath
// Trainee Engineer @ Eton Solutions
📍 Bengaluru, Karnataka, India
150+
PRs Reviewed
20+
Tool Users
50+
Microsvcs
Agentic AI LLM Infra RAG Azure C#/.NET Python
⭐ Leadership Recognition
Recognized by Director of Engineering & Chief Solution Architect for production AI tooling, agentic workflows, and platform impact.
Currently Building
AI Companion — a psychological partner that lives in your daily conversations. Not an assistant. A presence.
Scroll

Who I am

Engineer at heart.
Builder by nature.

I'm an AI Platform & Developer Tools Engineer who builds production systems, not demos. I specialize in the intersection of agentic AI, developer tooling, and distributed FinTech infrastructure — systems where architecture, reliability, and scale actually matter.

Currently at Eton Solutions as a Trainee Engineer, I build internal AI tooling for engineering workflows — a production serverless PR review agent, Eton Dev used by 20+ engineering and product users, and an active R&D 6-agent code pipeline backed by a dual-layer RAG system.

On the infrastructure side, I own three production financial microservices inside a 50+ microservice event-driven architecture, and I have worked on modernization from older stored-procedure-driven workflows toward event-driven service flows. I resolved a critical processing bottleneck from 15–20 hours to under 15 minutes across the distributed system, with recognition from senior engineering leadership for the AI platform work.

Core Stack
PythonC# / .NETJavaScript Claude / GPT-4 / MistralAzure AIFastAPI QdrantLlamaIndexLangFuse Azure FunctionsEvent-Driven ArchRAG VSIX / VS SDKMSAL.NETSQL
15→15m
Performance Impact
100K+ records: 15–20 hrs → 15 min
150+
PRs Reviewed Autonomously
Org-wide in 30 days, production
6
Agent Pipeline (Eton ARC)
Triage → Discovery → Council → Opus
8.62
CGPA — B.E. CSE, 2025
PES Institute of Technology

Where I've worked

Experience

Production engineering at a FinTech product company — building AI infrastructure, microservices, and developer tooling that ships.

Jul 2025 — Present Current
AI Platform & Developer Tools Engineer — Trainee
Eton Solutions — Wealth Management Platform
  • Built internal AI platform tooling for engineering workflows — a production autonomous PR reviewer, Eton Dev used by 20+ engineering/product users, and an active R&D 6-agent code pipeline. Recognized by Director of Engineering & Chief Solution Architect for LLM and agentic AI work.
  • Resolved a critical EDA bottleneck: 15–20 hours → 15 minutes across a 50+ microservice distributed system — a 98% latency reduction with zero regression.
  • End-to-end ownership of three production financial microservices inside the wider 50+ service EDA platform: EliminationService (trade logic), DFRulesProcessorService (rules engine), JournalEntryPersistService (ledger persistence).
  • Worked on modernization from older stored-procedure-driven financial processing toward event-driven microservice flows, including routing, queue-based processing, and service-boundary reasoning.
  • Built a serverless PR Review Agent on Azure Functions that autonomously reviewed 150+ pull requests org-wide — with delta re-reviews, @agent command system, and LangFuse observability.
  • Built Eton Dev — a VS 2022 VSIX extension embedding multi-provider LLM routing (Claude, Mistral, GPT-4), code review, auto-fix pipeline, and full Azure DevOps PR lifecycle in one panel for internal teams.
  • Designed Eton ARC — an active R&D 6-agent hierarchical code pipeline (Triage → Discovery → Specialist Council → Moderator → Opus → Execute) backed by a dual-layer Qdrant + LlamaIndex RAG system.
  • Currently building CI/CD workflow automation and agentic engineering systems that help senior developers debug, reason about services, and accelerate delivery.
PythonC# / .NETAzure Functions Claude / GPT-4 / MistralQdrantLlamaIndex LangFuseVSIX / WebView2MSAL.NET Azure DevOpsFastAPIEDA
Jul 2024 — Jun 2025 Internship
Full-Stack Developer Intern
Spurzee Technologies — FinTech & Trading Systems
  • Built a real-time stock analytics platform with live candlestick visualization, SEBI market data APIs, and 20+ automated pattern detection algorithms.
  • Integrated LLM-assisted trade signal generation and automated options execution — improving data processing throughput by 25%.
  • Deployed a Random Forest regression model for stock price prediction — full lifecycle from training through production on AWS and DigitalOcean.
PythonReactNode.js FlaskAWSDigitalOceanSQL
2025Invited
Alumni Mentor — Agentic AI & Developer Tooling
PES Institute of Technology and Management
  • Invited back to campus to conduct sessions on practical AI platform engineering — multi-agent system design, RAG architecture, LLM infrastructure, and the gap between academic ML and production AI.
  • Mentored juniors on agent orchestration patterns, prompt engineering discipline, and how AI-assisted developer workflows are structured in industry.

Proof signals

Delivered work, not demos.

Sanitized, recruiter-safe signals from production engineering and internal AI platform work.

150+
PRs Reviewed
Production Azure DevOps review agent in org-wide use.
20+
Internal Users
Engineering/product users on Eton Dev AI workflows.
98%
Latency Reduced
EDA bottleneck reduced from 15-20 hours to under 15 minutes.
3
Owned Services
Production financial microservices owned end-to-end.
11
Indexed Services
Active R&D code intelligence pipeline over service context.

What I've built

Engineering Projects

🤖
Production · Serverless
PR Review Agent — Autonomous Code Review Platform
Serverless dual-function architecture on Azure Functions — HTTP trigger acknowledges Azure DevOps webhook in <1 second; full LLM review runs async in queue-triggered processor with 9-minute budget. Handles 150+ PRs org-wide with context-aware delta re-reviews, @agent command interface, and LangFuse observability tracing every LLM call. 900-line system prompt across 7 review dimensions with confidence-based finding downgrade and drop-in-ready code fixes.
✦ 150+ PRs reviewed autonomously in 30 days
Architecture
HTTP trigger → Storage Queue → queue processor, keeping Azure DevOps webhooks under 1s while LLM review runs async.
Hard Part
Resolved Azure DevOps thread IDs across payload variants, with REST fallback to keep @agent replies nested correctly.
Reliability
Commit-keyed dedup, event normalization, confidence downgrades, and LangFuse traces for every LLM review path.
Developer UX
@agent commands for re-review, skip, focus, context, explain, and help directly inside PR threads.
PythonAzure FunctionsAzure DevOps Mistral / Claude / GPT-4LangFuseFastAPI
PR Review Agent dashboard PR Review Agent review output
Add PR Review screenshots
🧠
Production · Internal Platform · 20+ Users
Eton Dev — AI Development Companion (VS Extension)
VSIX extension for Visual Studio 2022 used by 20+ engineering and product users — AI chat, pre-PR code review, bug investigation, two-phase auto-fix engine, and full Azure DevOps PR lifecycle in one panel. Multi-provider LLM routing (Claude, Mistral, GPT-4), silent MSAL WAM broker auth with AAD auto-discovery, undocumented TFS GUID resolution.
Adoption
Used internally by 20+ engineering/product users for review, investigation, PR, and AI-assisted workflows.
LLM Routing
Protocol-based provider layer routes between Claude, Mistral, GPT/Azure OpenAI, and custom backends by repo context.
Auto-Fix Safety
Planner validates scope within ±3 lines; patch generator uses anchored old/new blocks with atomic writes and backups.
Enterprise Auth
MSAL WAM broker, browser SSO fallback, JIT JWT caching, and Azure DevOps IdentityPicker GUID resolution.
C# / .NETWebView2Python MSAL.NETAzure AIVS SDK
Eton Dev VS Extension main view Eton Dev Visual Studio integration Eton Dev Orbit feature Eton Dev investigation tool
Add VS Extension screenshots
⚡
Active R&D · Multi-Agent
Eton ARC — 6-Agent Code Intelligence Pipeline
Hierarchical AI pipeline: Triage (Haiku) → grep-first Discovery → parallel Specialist Council (Sonnet) → Moderator → Opus Principal Review → Execute. Takes a Jira ticket, produces verified file-level diffs and xUnit test stubs ready for Claude Code. Dual-layer RAG: Qdrant vector DB (80-line overlapping chunks, 384-dim embeddings, 11 services) + LlamaIndex retrieval pipeline. Opus issues GO / CONDITIONAL-GO / NO-GO with full AC coverage verification.
✦ Tested on live tickets · 11 services indexed
Status
Active R&D system being refined for story/issue tickets, with strong code-grounded answers and developer review gates.
Retrieval
Dual-layer Qdrant/LlamaIndex RAG: source chunks plus service profiles across 11 services.
Safety
Opus produces GO / CONDITIONAL-GO / NO-GO with AC coverage, evidence, regression risk, and scope-creep removal.
Output
Generates exact file operations, PR descriptions, build order, and xUnit stubs for Claude Code/Codex handoff.
Claude Opus 4.5Sonnet / HaikuQdrant LlamaIndexFastAPIPython
Eton ARC architecture overview Eton ARC agent workflow Eton ARC pipeline details Eton ARC system diagram
Add pipeline architecture images
🔁
Production · Backend Platform
EDA Latency Fix — SP-Driven Workflow to Event-Driven Services
Diagnosed and resolved a core routing bottleneck in a 50+ service event-driven financial platform, modernizing older stored-procedure-driven processing into service-owned flows with queue-based execution, idempotency, structured logging, and Dapper-backed persistence.
✦ 100K+ records · 15–20 hours → under 15 minutes
Problem
Legacy financial processing created long-running batches and delayed downstream reporting/analytics.
System
Bank feed ingestion → validation → enrichment → journal entry creation → MS SQL persistence.
Reliability
Idempotent service flows, retry-safe processing, structured error handling, Azure Monitor/Logs, and CI/CD release discipline.
Ownership
Owned EliminationService, DFRulesProcessorService, and JournalEntryPersistService end-to-end.
C# / .NETDapper DALMS SQL Azure Event GridAzure MonitorOAuth/JWT IdempotencyMicroservices
⚛️
Research · Published Paper
Insurance Risk via Quantum Computing
Published IJIRT paper on insurance risk prediction using a hybrid quantum-classical approach with QSVM/Qiskit concepts, Flask-based insurance workflows, batch processing, and risk dashboards.
IJIRT 171735QiskitIBM QuantumPythonFlask
Read Published Paper ↗
📈
Research · ML
LSTM + Sentiment Analysis — Stock Forecasting
LSTM + financial news NLP pipeline benchmarked against 5 regression models. 15% forecast accuracy improvement.
PyTorchLSTMNLPscikit-learn

How I think

Sanitized architecture patterns

High-level patterns only. Company internals, client data, service names, repos, endpoints, and proprietary workflows are intentionally omitted.

Async AI Review Pipeline
Webhook→ Queue→ LLM Review→ PR Feedback

Designed to acknowledge source-control events quickly while long-running AI work happens safely in the background.

IDE-Native AI Tooling
VSIX→ Local Backend→ Provider Router→ DevOps APIs

Keeps review, debugging, AI chat, auto-fix, and PR workflows inside the developer's existing environment.

Ticket-To-Code Agent Flow
Ticket→ Discovery→ Specialists→ Review Gate→ Execution Plan

Uses deterministic code search first, then retrieval and agent review to keep plans grounded in actual source context.

Legacy Flow Modernization
Batch/SP Flow→ Events→ Idempotent Services→ Observable Delivery

Moves long-running financial processing toward retry-safe, traceable, service-owned event flows.

Toolbox

Technical Skills

AI & Agentic Systems
Multi-agent orchestrationLLM infrastructure GenAI / RAGLlamaIndexQdrant LangFusePrompt engineering AI observabilityMCP Tool callingVector databases
Languages & Backend
PythonC# / .NETJavaScript JavaC/C++SQL FastAPIFlaskNode.js REST APIsDapper
Cloud & DevOps
Azure FunctionsAzure AI Azure DevOps REST APIAzure Event Grid AWSDigitalOcean DockerGitHub Actions CI/CDVSIX / MSBuild
Architecture & Infra
Event-driven architectureMicroservices Distributed systemsMessage queues MSAL.NET / OAuth 2.0WebView2 IdempotencyRetry-safe workflows Dapper DALAzure Monitor / Logs VS SDK (WPF, VSCT)PyTorch sentence-transformersMS SQL Server
🤖
Ask Deepak's AI
Projects, impact, skills, or role fit.
AI
Deepak's AI
Online — trained on Deepak's portfolio
AI🤖
Hi, I can answer questions about Deepak's production engineering work, AI tooling, microservices ownership, and project impact. Ask about his projects, skills, experience, or fit for a role.
What role would Deepak be best for? Show the PR review architecture What production systems has he shipped? Explain the agent orchestration 📬 Send a message to Deepak

Let's connect

Get In Touch

Open to great
opportunities

AI platform roles, developer tooling, agentic systems, backend infrastructure — if it's technically challenging and impactful, I'm interested. Open to remote and relocation.

📄
Resume
Open PDF
✉️
Email
khimavathdeepak@gmail.com
📞
Phone
+91 7204209095
💬
WhatsApp
Message directly
💼
LinkedIn
deepak-khimavath-338074219
🐙
GitHub
DeepakkhimavathBB
Send a message