0%
SPSPSathwik Prathi

Lead AI Engineer · Circuitry.ai

I build systems thatthink at scale.

4+ years building production LLM, RAG and agentic systems. I design multi-agent workflows that reason, call tools and decide autonomously — plus the reliability layer that makes it safe.

Scroll
Agentic AIMulti-Agent OrchestrationA2AMCPAutonomous DecisioningRAG SystemsLLM ReasoningEnterprise ReliabilityFastAPIKubernetesLLM SystemsAgentic AIMulti-Agent OrchestrationA2AMCPAutonomous DecisioningRAG SystemsLLM ReasoningEnterprise ReliabilityFastAPIKubernetesLLM Systems

— / What it looks like

A conversation with the systems I build

agent-session · claims-platform
you

Why was claim #48213 auto-approved instead of escalated?

Scoring claim against coverage policy…
Calling MCP tool → get_claim_history(claim_id)
Checking anomaly-detection thresholds…
agent

Trace:coverage_check ✓anomaly_check ✓payment.release()

01 / Selected Work

Six problems worth telling you about

01 / Lead

Agent-to-Agent Decision Platform

Specialized agents for intake, document understanding, coverage validation and anomaly detection — coordinated through supervisor-pattern handoffs and MCP tool calling, replacing a monolithic flow.

Agentic AIA2AMCP
02 / Lead

Enterprise Reliability Layer

Durable checkpointing so workflows survive crashes and redeploys, idempotency on every side effect, compensating actions for partial failure — what makes it safe to let an agent move money.

Distributed SystemsKubernetesReliability
03 / Senior

Agentic Reasoning Framework

Planning, reflection, dynamic tool selection and adaptive execution over complex enterprise queries — decomposed into parallel RAG workflows with reranking. Positive ratings went 95% → 99%.

RAGReasoningAgentic AI
04 / Senior

Claim Validation System

Cross-verifies images, video, descriptions, causes and corrective actions against each other to surface inconsistencies — cutting fraudulent claims and making automated decisions trustworthy.

Computer VisionFraud DetectionAI
05 / SE

Dialogue Intelligence Layer

Replaced a legacy Rasa system with a proprietary FastAPI layer, plus vision-based chunking that recovered tables and diagrams OCR was dropping. Costs down 60–70%, coverage past 99%.

FastAPILLMComputer Vision
06 / Personal

Eye Wellness Tracker

Real-time blink detection (MediaPipe, PyQt6) with an offline-first telemetry backend — SQLite outbox, async FastAPI, idempotent ingestion for unreliable networks.

MediaPipePyQt6FastAPI
Drag or scroll sideways→

02 / About

I build AI agentsthat make decisions —not justconversation.

At Circuitry.ai I lead the agents team behind our decision intelligence platform — multi-agent orchestration that scores a claim, then decides what happens next: release payment, retry a failed downstream call, or escalate to a human with the full reasoning trace attached.

0%cut in dialogue system costs
0%enterprise document coverage
0%latency reduction (10s → 5s)

03 / Experience

Where I’ve been

Click a role to see what actually shipped.

2017 — 2021B.Tech, Electrical Engineering IIT Dharwad

— / Stack

Toolbox

What I actually reach for. Highlighted items are where I spend most of my time.

ai & agents12
  • LLMs
  • RAG
  • Agentic AI
  • Reasoning Frameworks
  • MCP
  • A2A
  • LangGraph
  • LangChain
  • CrewAI
  • OpenAI API
  • Anthropic API
  • Vision-based Chunking
backend07
  • FastAPI
  • Django
  • DRF
  • Flask
  • SSE streaming
  • JWT
  • Distributed State Management
ml & cv07
  • PyTorch
  • Scikit-learn
  • BERT
  • OpenCV
  • MediaPipe
  • NLP
  • Semantic Chunking
infra08
  • AWS
  • GCP
  • Docker
  • Kubernetes
  • Temporal
  • Redis
  • OpenTelemetry
  • Datadog
data06
  • PostgreSQL
  • Milvus
  • Neo4j
  • Redis
  • Chroma
  • SQLite
currently03
  • agent reliability evals
  • production agentic systems
  • autonomous workflows

05 / Contact

Let's build agents
worth trusting.

Say hello →