BUILD MODEGHLet's build
AI Engineer
Agent Systems
Bangalore, India

Bhargav Kacharla

I design and ship production AI and backend systems: coding agents, tool-using workflows, APIs, workers, queues, context pipelines, evaluations, and reliable platform infrastructure.

AI AgentsBackend APIsCoding AgentsTool CallingContext EngineeringAgent EvalsTypeScript / Python
View featured work →GitHubContact↓ Download Resume
01 / SYSTEMS

AI and backend systems I can own end-to-end.

A fast scan of the agent, retrieval, evaluation, and production backend surfaces I build.

[01]

Agent orchestration

Planning loops, routing, tool execution, retries, multi-step workflows, and observable agent traces.

[02]

Context engineering

Prompt pipelines, memory, structured project context, tool results, and compact task state for grounded agent work.

[03]

RAG pipelines

Ingestion, chunking, retrieval, reranking, answer grounding, citation flow, and hallucination reduction.

[04]

Vector databases

Embedding strategy, metadata filters, semantic search, similarity tuning, and retrieval quality debugging.

[05]

Structured outputs

Typed schemas, tool-call contracts, JSON outputs, validation, parsing, and deterministic handoff to backend services.

[06]

Evals

Deterministic checks, LLM-as-judge scoring, known-bug benchmarks, precision/recall tracking, and regression suites.

[07]

Backend platforms

APIs, workers, queues, Postgres/MongoDB state, auth, webhooks, deployments, and production operations.

02 / FEATURED WORK

Agent systems built.

Agent Workflow Engine
Nunchi.trade · Agent Commerce Protocol Execution Layer
2025 — Present

Built the agent execution layer for Nunchi's Agent Commerce Protocol stack, with job discovery, on-chain handlers for oracle updates, liquidations, settlements, and payment settlement.

Designed the flow so an autonomous agent can discover a job, execute it, and finish settlement without a human stepping in.

Built the signer safety layer with destination allowlists, function selector allowlists, per-transaction value caps, and per-block rate limits.

Integrated the agent layer with the underlying perpetuals system and near-real-time indexing pipelines that scale on-chain ingestion to 1000 TPS.

LLM AgentsTool CallingERC-8183ERC-8084Safety ControlsRate LimitsIndexers1000 TPSBackend
RAG Knowledge Assistant
AI Engineering Project · Protocol Documentation Assistant
2025 — 2026

Built a TypeScript/Bun RAG pipeline for blockchain protocol documentation with document ingestion, section-aware chunking, Qdrant vector search, and citation-grounded LLM answers.

Implemented incremental indexing with content and chunk hashes, stale vector deletion, and a Postgres document registry to avoid unnecessary re-embedding.

Built retrieval and answer-quality evals with judge scoring for faithfulness, relevance, citation correctness, structured JSON validation, missing-context handling, and per-stage trace timings.

TypeScriptBunQdrantPostgresRAGJudge ScoringCitations
AI Smart-Contract Auditor
AI Engineering Project · Autonomous Hermes Auditor
2025 — 2026

Built an autonomous Hermes audit agent with honcho memory provider and scheduled jobs for live Code4rena/Sherlock bounty discovery and continuous exploit-knowledge ingestion.

Curated a DeFi exploit knowledge base from 650+ on-chain incidents and 331+ contest findings.

Supported audits across 14+ codebases, including 4 live bug bounties with $500K+ in prize pools.

HermesHoncho MemoryCode4renaSherlockExploit KBCron JobsAutonomous Agents
Backend Indexing & Analytics
Novastro · Real-Time Data Systems
2024 — 2025

Built in-house indexers for EVM chains using TypeScript, GraphQL, and MongoDB to stream, process, and query on-chain data for analytics.

Developed backend systems for DeFi protocols, including real-time data aggregation and analytics pipelines for trading metrics and liquidity tracking.

Worked across EVM, Move, and Solana environments, building systems for data processing and execution logic.

TypeScriptGraphQLMongoDBEVMMoveSolanaAnalytics
Execution Infrastructure
Xalts · Distributed Backend Services
2023 — 2024

Implemented account abstraction and backend infrastructure for transaction execution pipelines.

Built internal services to coordinate execution across bundlers and relayers with job scheduling, retries, and transaction batching.

Increased transaction processing throughput from 100 TPS to 1000 TPS.

ERC-4337BundlersRelayersJob SchedulingRetriesBatching1000 TPS
03 / AGENTS

Agents backed by real platform engineering.

This portfolio is for applied AI engineering: systems where agents use tools, preserve context, call backend services, make changes, and verify outcomes. The focus is product-grade agent behavior, not ML research.

01Coding agents
Repo-aware workflows that inspect files, propose plans, edit code, run tests, and summarize diffs in a reviewable way.
02Tool-using agents
Agents connected to APIs, MCP servers, browsers, databases, logs, background jobs, and internal systems through typed tools.
03Agent products
Backend-backed workflow surfaces for support, engineering operations, audit prep, debugging, and repetitive technical work.
04 / RELIABILITY

Agents need engineering discipline.

Useful AI systems are not just prompts. They need bounded tool access, good context, evals, rollback paths, observability, and clear human control points.

✓Design tool permissions, approval gates, and traceable execution paths before letting agents mutate real systems.
✓Build eval suites for task success, tool-call correctness, regression behavior, and quality of final outputs.
✓Use structured context, repo maps, summaries, and retrieval so agents know what matters before they act.
✓Instrument long-running sessions with logs, checkpoints, retries, and clear recovery behavior.
05 / TIMELINE

AI engineering arc.

Nunchi.trade · Senior Engineer (AI Systems)

2025 — Present

Built the AI-native execution layer for DeFi protocols, agent marketplace, perpetuals execution environment, and indexing pipelines that scale on-chain ingestion to 1000 TPS while keeping agent jobs near real time.

Novastro · Backend Engineer

2024 — 2025

Built EVM indexers, GraphQL/MongoDB data systems, DeFi analytics pipelines, trading/liquidity tracking, and cross-environment backend work across EVM, Move, and Solana.

Xalts · Backend Engineer

2023 — 2024

Implemented account abstraction and transaction execution infrastructure across bundlers and relayers, with job scheduling, retries, batching, and throughput improvements from 100 TPS to 1000 TPS.

Oddz · Blockchain Engineer

2021 — 2023

Led smart contract development for the Oddz options protocol and built automation systems for expiry handling, staking rewards, and dynamic risk adjustments with 100% production uptime.

Rakuten · Data Engineer

2020 — 2021

Architected data lake infrastructure for large-scale telecom data using MinIO and YugaByte, supporting data science and analytics workflows.

OpenText / TCS · Software Engineer

2015 — 2020

Built enterprise REST services, web applications, third-party API integrations, SQL Server systems, and backend software for banking, telecom, and content-management platforms.

CONTACT

Building an agent product or AI backend?

Reach out for agent architecture, coding-agent workflows, backend APIs, workers, queues, context systems, eval design, or production AI integration.

Email meGitHub ↗X ↗LinkedIn ↗