Keshav Bathla

AI Systems Consultant · Senior Software Engineer

I build production AI agents and high-scale backend systems.

I help founders and engineering teams turn AI prototypes and complex platform requirements into reliable systems—with secure tools, measurable evals, human approvals, and production-grade infrastructure.

Remote worldwide & in-person in Bengaluru · India, US, Europe & Middle East · Harness · Ex-CoinSwitch · Ex-Blinkit

Explore Consulting View Selected Systems Technical leadership résumé
  • 80% On-Call Ramp-up Cut
  • 95% CVE Patching Effort Cut
  • 50K Orders/Min · <200ms p99
  • 20M+ Users at Platform Scale

Work

Selected systems

A selection of systems where architecture, reliability, and measurable outcomes mattered.

01 · Harness · AI Engineering Systems

From operational context to reviewed pull requests

  • 80% On-call ramp-up cut
  • 95% CVE effort cut
  • Days → minutes Rollout wait

An internal developer agent that understands repositories, connects engineering context, and executes approved changes through a sandboxed CI workflow.

Challenge
Debugging and routine customer fixes required repeated context gathering and on-call involvement.
System
Request → repository analysis → proposed plan → human approval → sandboxed pipeline → tests → pull request. Built with Go, Slack, LSP, Jira, Confluence, Grafana, Harness APIs, PostgreSQL, and MCP tools.
Outcome
Support teams can diagnose and address customer issues from Slack with less on-call dependency.

Go · Slack · MCP · PostgreSQL · Grafana · Harness APIs

02 · CoinSwitch · Trading & Payments

Low-latency financial systems at 20M-user scale

  • $30M+ Liquidity
  • 50K/min Orders
  • <200ms p99 latency

Pricing, hedging, order routing, and derivatives systems built for low latency and sustained reliability.

Challenge
Serve millions of traders through volatile markets without depending on third-party market makers or brittle payment paths.
System
Four Python market-making services, a Go derivatives engine, and configuration-driven payments across crypto and other asset classes.
Outcome
$30M+ liquidity delivered, ~50,000 orders/minute, sub-200ms p99 on critical paths, and 99.999% trading availability. Figures come from related systems, not a single component.

Python · Go · PostgreSQL · Redis · Kafka · Kubernetes

03 · Blinkit · Product Discovery

10x faster product listing for 20M+ users

  • 1,000 → <100ms Latency
  • 20M+ Users

Modernized Product Listing and Discovery services while traffic and product complexity were increasing.

Challenge
Legacy listing paths were approaching ~1,000ms while consumer traffic scaled toward 20M users.
System
Python 2/Flask → Python 3/FastAPI, async request processing, and monitoring through Grafana and New Relic.
Outcome
Listing latency reduced from approximately 1,000ms to under 100ms at 20M+ consumer scale.

Python · FastAPI · Grafana · New Relic

Expertise

AI & platform systems I work on

Agentic Engineering Systems

Production agents that plan, call tools, and wait for human approval before changing code or infrastructure. The loop is the product: context in, reviewed change out.

LangGraph · MCP · OpenAI · Claude

Retrieval, Evals & Guardrails

Grounded answers need retrieval, regression evals, and output constraints—not just a prompt. I build RAG pipelines and eval harnesses for systems that fail loudly.

RAG · hybrid search · evals · structured output

Distributed Backend Systems

High-throughput services, event streams, and low-latency APIs for products at million-user scale. Trading, payments, and discovery taught the same lesson: p99 is the spec.

Python · Go · TypeScript · Kafka · SQS · Redis

Platform Reliability & Observability

Capacity, latency, and incident response treated as product features. If you cannot see the path, you cannot ship the agent that walks it.

Kubernetes · PostgreSQL · Grafana · Prometheus · New Relic

Consulting

Need senior help with a production AI system?

I offer bounded engagements for production AI agents, secure enterprise MCP integrations, RAG evaluation, and high-scale backend architecture.

Remote worldwide & in-person in Bengaluru · India, US, Europe & Middle East

Review Consulting Offers Discuss a Project

Career

Experience

Download Résumé

  1. Harness

    Senior Software Engineer · Bengaluru

    • CodePulse and Plan-to-PR: Built an internal developer agent (Go Slack bot & CLI, Go/Java LSP) that turns repository and operational context into approved change plans and tested pull requests via sandboxed CI.
    • Incident diagnosis: Built an on-call triage agent that correlates PagerDuty alerts, deployments, Grafana dashboards, and traces, cutting new-engineer on-call ramp-up time by 80%.
    • Security remediation: Built an automated vulnerability bot tracking critical/high CVEs, patching dependencies, verifying CI builds, and raising PRs, reducing manual remediation effort by 95%.
    • Feature rollouts: Built Slack workflows for PMs to safely trigger feature flags, canary deployments, and experiments, reducing rollout wait times from days to minutes.
    • MCP infrastructure: Built production Model Context Protocol (MCP) servers and CLI integrations so tools like Cursor and Claude Code can safely run pipelines, query code, and check observability metrics.

  2. CoinSwitch

    Senior Software Engineer, Trade · Apr 2023 – Dec 2025

    • Built in-house market-making platform (4 Python microservices) providing $30M+ liquidity with 99.99% uptime, cutting third-party market maker dependency.
    • Built Go derivatives matching engine processing up to 50,000 orders/minute in high volatility with sub-200ms p99 latency and 99.999% uptime.
    • Launched real-time Portfolio Analytics service for 20M users, optimizing Redis caching and DB queries to maintain ~400ms latency during market spikes.
    • Scaled 10+ critical microservices (Trade, Rates, Auth, Portfolio) to handle 1M concurrent users at peak with zero downtime.

    Senior Software Engineer, Payments · Oct 2022 – Mar 2023

    • Built configuration-driven Go payment gateway plugin, reducing new gateway onboarding time to under 10 minutes while processing 1,000+ tx/minute.
    • Engineered unified payment infrastructure across multiple asset classes: Crypto, Mutual Funds, and US Stocks.

    Software Engineer · Jan 2022 – Sep 2022

    • Rewrote core payment service architecture for multi-asset investing, scaling to millions of transactions per day across UPI, NetBanking, and VA transfers.

  3. Blinkit

    Software Engineer · Intern → Full-time · Gurugram

    • Migrated Product Listing & Discovery services from legacy Python 2/Flask to Python 3/FastAPI, scaling to 20M+ users.
    • Reduced Product Listing API latency from ~1,000ms to under 100ms with async processing and New Relic/Grafana observability.
    • Engineered Instant Replacement feature for fresh produce on customer app, removing critical checkout bottlenecks.

  4. Oppia Foundation

    GSoC Student Developer · Remote · Nov 2018 – Jun 2019

    • Selected for Google Summer of Code 2019; cut educational page rendering speed by 2 seconds across Oppia.
    • Successfully merged 10+ pull requests spanning AngularJS migration, Python backend optimizations, and safe dependency upgrades.

Writing

Featured writing

Off-hours

The training log

Same operating system as the engineering: measure, iterate, don’t skip the boring work. 485 km in 2026, Wipro Bangalore Half Marathon (21.75 km), 6:19 min/km PR, PPL split at 67 kg.

Logged in public

Race milestones, monthly mileage breakdown, a PPL cut to 67 kg, and daily fuel stack — the same discipline as building systems.

Half marathon · 6:19 /km PR · PPL · Bengaluru

Open the athletic log

Contact

Three ways to work with me

Pick the path that matches what you need — I reply personally within 24 hours.

Path 1

Production AI Consulting Sprint

Agent architecture & MCP integrations for funded teams — bounded 2–4 week engagements.

Discuss a Consulting Sprint

Path 2

Freelance AI Engineering

Scoped LLM & agent builds — smaller and more flexible than a full consulting sprint.

Discuss a Freelance Build

Path 3

Full-time Senior AI Engineer / FDE

Bengaluru or remote — Senior AI Engineer or Forward-Deployed Engineer (FDE) roles.

Share an FTE / FDE Role
Book a 30-Min Architecture Fit Call Review Résumé

Prefer a diagnostic first? Start with the RAG Pipeline & Eval Audit

WhatsApp and Indian mobile numbers are welcome. Fit calls default to IST business hours unless you name another timezone.

I reply personally within 24 hours · Direct builder access