Cyprien Dcunha — Senior Software Engineer
Open to new senior roles. Based in Hyderabad, India; open to roles in Bangalore, London, Dublin, Singapore, or remote. Contact: dcunha.cyprien@gmail.com.
Senior software engineer with roughly eight years of production experience, the bulk of it — nearly six years at Amazon (FAANG) Finance Automation, promoted from SDE I to SDE II on merit. Now a Senior Software Engineer at Salesforce Hyperforce AI. Specialises in distributed systems, event-driven architecture, and GenAI automation. Equally comfortable designing event-driven microservices, building agentic AI workflows, and shipping production LLM systems at scale. Open-source work includes a LoRA fine-tuning pipeline for local LLMs (VoiceTune) and from-scratch neural-network implementations (Hello NN).
Ideal roles
Open to senior or staff-level backend / distributed-systems roles and applied-GenAI / LLM engineering roles. Background: nearly six years at Amazon Finance Automation, promoted from SDE I to SDE II, now Senior Software Engineer at Salesforce Hyperforce AI. Details in experience and projects below. More at cyprien.in and the Resume.
Why hire Cyprien Dcunha?
- Built harness plugins, MCPs, and autonomous Agent Skills at Salesforce to accelerate developer productivity, consolidating integrations with quality, accessibility, and provenance controls, achieving almost 100% weekly AI engagement and cutting PR review wait times by ~50%.
- Co-lead architect of the vendor-holdbacks re-architecture at Amazon — decomposed a batch SQL monolith into an event-driven microservice platform, moving payment processing from overnight batches to near real-time.
- Shipped a production LLM-driven financial reconciliation engine — end-to-end RAG pipeline over historical analyst notes, semantic mapping of free-text fields, human-in-the-loop review.
- Built a self-service rule engine that compressed rule authoring from weeks of engineering work to minutes and now serves thousands of production rules authored directly by business owners.
- ML-driven risk model for seasonal holdback decisions that freed analysts from manual per-vendor triage through Prime Day, Black Friday, and holiday peaks.
- Lead on Salesforce's first AWS-to-GCP service uplifts — engineered a multi-substrate routing SDK, stood up GCP infrastructure, ran production readiness reviews, and migrated live customer workloads across 9 Core, Platform, and Agentforce services.
- Multiple engineering recognition awards at Amazon; IBM Call for Code prize winner; top-three TopCoder finish; top-of-cohort CS undergraduate at DJSCE Mumbai.
Specialisations
Distributed systems & event-driven architecture: microservices, domain-driven design, idempotent message processing, SQS / SNS / EventBridge, DynamoDB, CloudFormation / CDK, Java on AWS Lambda and ECS.
Applied Generative AI & LLM systems: retrieval-augmented generation (RAG), LLM-in-the-loop workflows, AWS Bedrock, AWS AgentCore, LangChain, prompt engineering, agent memory, vector retrieval — shipped in a regulated, high-volume finance setting.
Multi-cloud / platform engineering: AWS (Lambda, ECS, Step Functions, SQS, SNS, EventBridge, DynamoDB, RDS, S3, Redshift) and GCP (GKE, Cloud SQL), with Kubernetes, Docker, Jenkins, and Spinnaker across CI/CD.
Backend & systems programming: production Java and Python; earlier C++ low-latency work as a student intern.
Experience
- Salesforce — Senior Software Engineer, Hyperforce AI. Nov 2025 – present. Hyderabad. Building Agentic SDLC tooling (harness plugins, MCPs, autonomous Agent Skills) to accelerate developer productivity. Also lead on the Multi-Substrate Infrastructure (Core, Platform & Agentforce GCP Uplift) programme (9 services uplifted with dynamic substrate routing SDK).
- Amazon (FAANG) — Software Development Engineer II, Finance Automation. Dec 2021 – Nov 2025. Hyderabad. Promoted from SDE I. Co-lead architect on vendor-holdbacks re-architecture. LLM-driven reconciliation engine. Self-service rule engine. Policy-based approvals / reminders email orchestrator.
- Amazon (FAANG) — Software Development Engineer I, Finance Automation. May 2019 – Dec 2021. Hyderabad. Seasonal holdbacks ML risk model. Special-merchant data segregation across views, APIs, and legacy reports.
- IRageCapital - High-Frequency Trading — C++ Development Intern. Jun 2017 – Aug 2017. Mumbai. C++17 HFT market-replay simulator with lock-free hot-path data structures.
Open-source AI / ML projects
- SystemDesign.ai — desktop app for AI-assisted system design mock interviews with real-time audio and Excalidraw canvas diff tracking. Electron · React · LLMs · TypeScript (GitHub).
- CP Solutions — search and explorer for competitive programming solutions across LeetCode, Codeforces, HackerRank, and CodeChef. C++ · Web (GitHub).
- VoiceTune — end-to-end LoRA fine-tuning pipeline for local LLMs, turning personal audio call recordings into fine-tuning datasets and running Gemma fine-tunes on GCP. Python · LoRA · Gemma · GCP.
- Hello NN — from-scratch implementations of feedforward, CNN, RNN, autoencoders, Hopfield, and Restricted Boltzmann Machine networks. Python · Jupyter · TensorFlow · NumPy.
- Full list at github.com/ccd97.
Education
Bachelor of Engineering, Computer Science — D. J. Sanghvi College of Engineering, University of Mumbai (2015–2019). Graduated top-of-cohort.
Skills
Languages: Python, Java, C++, TypeScript, JavaScript, SQL.
Backend & architecture: distributed systems, agentic systems, microservices, event-driven architecture, domain-driven design, REST APIs, Spring, Guice / Dagger.
AWS: Lambda, ECS, Step Functions, EventBridge, SQS / SNS, DynamoDB, RDS, S3, Redshift, Bedrock, AgentCore.
GCP & cloud-native: GKE, Cloud SQL, Docker, Kubernetes, Jenkins, Spinnaker.
Generative AI & ML: LLMs, RAG, agent memory, prompt engineering, LangChain, AWS Bedrock, AWS AgentCore, TensorFlow, NumPy, Pandas.
Links