Looking for a job

Abhik Maiti

@abhikmaiti

Message

Agentic AI Developer at Scholastic who built an enterprise MCP platform used weekly by 350+ engineers across 23 tools.

India

Message

At Scholastic, I lead the org-wide agentic intelligence platform used weekly by 350+ engineers. I designed and built RBAC-scoped MCP servers for 23 engineering and security tools behind a unified JSON-RPC interface.

I also engineered a real-time correlation agent linking vulnerability scans and tracking, cutting mean time to identify vulnerabilities by 70%. A routing topology across 10 specialized agents reduced cross-tool incident lookup latency to under 1.8 seconds at p95.

At PRM Fincon, I served as technical owner of a financial document-intelligence platform for 15+ institutional clients. I evolved it from CNN and OCR pipelines into a multimodal agentic RAG and NL-to-SQL architecture, and led a 15-person cross-functional team.

Earlier, at Infrrd.ai, I developed CNN-based image-matching systems and a model for automated PDF form layout extraction. At IIT Kharagpur, my research included photogrammetric 3D reconstruction and probabilistic modelling; I hold a PhD and have published seven peer-reviewed papers.

Experience

Work history, roles, and key accomplishments

Current

Agentic AI Developer (Enterprise AI Tech Lead)

Current

Scholastic

Oct 2025 - Present (1 year)

Tech lead for Scholastic's org-wide agentic intelligence platform, an automated interface across 23 engineering and security tools used weekly by 350+ engineers. Designed RBAC-scoped MCP servers, reduced MTTI by 70%, and cut cross-tool incident lookup latency to under 1.8s.

MCP Multi agent orchestration Mastra LangGraph JSON RPC RBAC Datadog Qualys SonarQube Jira Gemini Flash Semantic Caching (GPTCache)LLM As A Judge Evaluation Frameworks

Data Scientist (Technical Lead, AI Platforms)

PRM Fincon

Jun 2019 - Jun 2026 (7 years)

Technical owner of a financial document-intelligence platform for 7 years, directing a 15-person team. Tripled ingestion throughput, reached 94% structural extraction accuracy, and improved multi-hop question accuracy by 40%.

CNN OCR YOLO Qwen2 VL Donut Lang Chain Pgvector NL To SQL PostgreSQL Active Learning Data Annotation MobileNetV3 EfficientNetV2 RetinaNet Calamari Generative Transformer OCR

Senior Software Engineer

Infrrd.ai

Oct 2018 - Jun 2019 (8 months)

Developed a CNN-based system for cross-view matching between ground-level and overhead imagery. Built a Siamese CNN model for matching building images with Google Street View counterparts and designed an object detection-driven model for automated layout extraction from PDF forms.

CNN Siamese Networks Object Detection Layout Extraction Computer Vision

Senior Research Fellow

IIT Kharagpur

Sep 2014 - Jun 2018 (3 years 9 months)

Built the first probabilistic slope stability framework for opencast mining using photogrammetric 3D pit geometry. Developed 3D reconstruction-based numerical modelling pipelines and machine learning models for coal quality prediction.