I'm Shrijeeth S

█

Building production AI systems — 500K+ transactions, 97% accuracy, sub-50ms latency. From fine-tuning models to deploying at scale on GCP.

[ contact ][ view work ]

About

Years Exp

K+ Txn/Month

Accuracy

<0ms

Latency

AI/ML engineer with ~3 years shipping production LLM systems, multi-agent pipelines, and low-latency inference services. Comfortable working across the full stack — from fine-tuning models to deploying FastAPI services on GCP. Previously joined Knowledge Keeper as founding engineer, building core AI infrastructure from scratch.

B.E. Computer Science

Rajalakshmi Engineering College, Chennai

2019 - 2023 · Specialization in AI/ML

◆Deep Learning Specialization — DeepLearning.AI

Experience

Software Development Engineer (AI/ML)

Shopup · Bengaluru

Nov 2025 – Present

[ - ]

Built Generator-Critic multi-agent system for invoice parsing — 500K+ txn/month, 97% accuracy, 40% fewer false positives.
Designed credential injection layer preventing API keys from reaching LLM context — adopted across all internal tools.
Real-time ML pipelines via Temporal — sub-50ms latency, 30% higher throughput vs Airflow/Prefect.
Lead architecture across 5+ GCP projects with Docker, Arize Phoenix, MongoDB, PostgreSQL.
Built SilqClaw: AI agent orchestration platform automating full dev lifecycle across engineering.
Train interns on AI/ML practices; conduct Round 1 & 2 technical interviews for SDE candidates.

AI Engineer (Founding Engineer)

Knowledge Keeper · Remote

Aug 2024 – Nov 2025

[ + ]

Software Development Engineer (AI/ML)

Shopup · Bengaluru

Jun 2023 – Aug 2024

[ + ]

SDE Intern → ML Intern

Shopup / AMTDC · Bengaluru / Chennai

Dec 2021 – Jun 2023

[ + ]

Projects

[01]

CALLSIGN: SilqClaw
OBJECTIVE: AI agent orchestration platform: agent creation, tool assignment, skill management, and access control across the engineering team.
STATUS: ████████░░ DEPLOYED
RUNTIME: PythonFastAPIMulti-Agent+1
INTEL: classified

[02]↗

CALLSIGN: QueryGPT ADK
OBJECTIVE: Multi-agent NL-to-SQL with LLMs + Qdrant vector search. JWT auth, Redis rate limiting, Ollama + OpenAI backends.
STATUS: ██████░░░░ ACTIVE
RUNTIME: LLMsQdrantRedis+1
INTEL: ↗ github.com/Shrijeeth/QueryGPT-ADK

[03]

CALLSIGN: Compliance Monitor
OBJECTIVE: YOLO11 + Qwen3 VLMs on Kafka + FastAPI replacing 75% of manual compliance review in real-time.
STATUS: ████████░░ DEPLOYED
RUNTIME: YOLOVLMKafka+1
INTEL: classified

[04]↗

CALLSIGN: Phi-3 Fine-tuning
OBJECTIVE: Fine-tuned Phi-3-mini-128k on Text-to-SQL dataset using PEFT + LoRA for natural language to SQL generation.
STATUS: ████░░░░░░ PROTOTYPE
RUNTIME: PEFTLoRAPhi-3+1
INTEL: ↗ kaggle.com/code/shrijeethsuresh/text-to-sql-phi-3-4b-finetuning

[05]↗

CALLSIGN: Go Fiber Template
OBJECTIVE: Production-ready backend starter: OAuth, cloud storage, transactional email, async tasks via GoCraft + Redis.
STATUS: ██████░░░░ ACTIVE
RUNTIME: GolangFiberOAuth+1
INTEL: ↗ github.com/Shrijeeth/Golang-Fiber-App-Template

[06]

CALLSIGN: Doc Automation
OBJECTIVE: CrewAI + Qdrant + FastAPI multi-agent system for 10+ document types — 80% less manual processing.
STATUS: ████████░░ DEPLOYED
RUNTIME: CrewAIQdrantAgents
INTEL: classified