Backend, automation, and reliability for enterprise financial systems.
I’m Mayank Sahu, a Software Engineer at Broadridge (Hyderabad) with 3+ years of experience building scalable services, automation platforms, and reliability-focused systems. I’m currently building Continuum, an agent memory manager (STM → MTM → LTM) using embeddings and vector search to keep long-running agents consistent.
Now
I’m focused on engineering reliable backend services, automation platforms, and agent infrastructure that can sustain enterprise workloads.
✅ Automation platforms + CI/CD (Jenkins)
✅ Reliability improvements (observability + latency)
✅ Agent Memory Manager (STM → MTM → LTM)
Skills Toolkit I trust in delivery
Current focus: memory routing for long‑running agents (recency vs task relevance vs durable knowledge).
I care about reliability: debuggability, CI feedback loops, and measurable quality signals.
Delivery focus: scalable services, observability, and production-grade rollouts.
Experience Impact & growth milestones
- Designed and delivered internal engineering platforms and workflow automation for enterprise financial systems, improving release throughput by 40% and reducing manual operational effort by 60%.
- Built and maintained backend services (Java/Python) and integrations to support quality signals, execution workflows, and platform insights consumed by cross-functional teams.
- Implemented and optimized REST APIs and service-to-service communication patterns, improving scalability by 3× and reducing response latency by 45%.
- Strengthened reliability through CI/CD hardening (Jenkins, Docker) and observability improvements, reducing MTTD by 50%.
- Developed reusable engineering components and automation-driven workflows to standardize releases and improve developer productivity across multiple teams.
- Led code reviews and introduced maintainable design patterns for Java and Python services, improving quality metrics by 35% and reducing production issues by 25%.
- Built CI/CD pipelines with Jenkins, reducing deployment time by 70% and improving environment consistency.
- Partnered with stakeholders to define actionable platform insights (execution signals, reliability indicators) and shipped incremental improvements with clear ownership.
- Built automation-driven tooling and scripts for critical business workflows, improving repeatability and accelerating release readiness for enterprise applications.
- Re-engineered a legacy batch processing workflow in Java and Python, improving throughput by 5× and reducing processing time from hours to minutes.
- Built full‑stack features (React + Node.js + MongoDB) for 500+ corporate clients.
- Developed AI invoice matching (Python + scikit‑learn), reducing manual processing by 80%.
- Delivered 12 features across 3 releases using Agile sprints.
Projects Hands-on buildouts
- Built a multi-layer memory system (STM/MTM/LTM) to manage conversation context and task history with lifecycle-aware retention and retrieval.
- Implemented embeddings + vector similarity search with a pluggable vector-store design to support multiple backends.
- Designed memory routing to query the right layer based on recency, task relevance, and durability, improving long-running agent consistency beyond basic RAG.
- Architected modular services (embedding, memory layers, retrieval/orchestrator) to keep providers and storage backends swappable.
- v1 delivered with STM and MTM complete; building v2 roadmap including memory UI (browse/edit/delete) and graph-style navigation.
Education Academic foundation
- Relevant Coursework: Data Structures and Algorithms, Machine Learning, DBMS, Software Engineering, Cloud Computing, Operating Systems, Computer Networks
- Leadership: Technical Lead at College Coding Club; organized technical workshops and peer learning sessions
- Activities: National Service Scheme (NSS) volunteer