About
Backend and Generative AI systems architect.
I am Suraj Pal. I design and build backend systems and Generative AI infrastructure that runs in production. I have worked in the Generative AI domain since 2023, and my work spans distributed services, API design, and the move from simple model integrations to autonomous agentic systems with persistent state. I build mostly in Go and Python on AWS, and today I lead the engineering behind Aedeon.
I care about systems that are fast, reliable, and simple to operate. I measure before I optimize, I keep interfaces small, and I write the reasoning down so the next engineer can follow it.
Experience
Aedeon
RemoteVP of Engineering
Dec 2025 to Present- Own the architecture the entire Aedeon product is built on, and lead its continued evolution.
- Lead a team of 7 engineers across the platform, agent, and infrastructure workstreams.
- Designed a code intelligence platform that ingests codebases into a Neo4j code graph and runs automated, security-focused pull request reviews with Codex and Claude Opus.
- Resolved Temporal heartbeat timeouts caused by unbounded Neo4j dependency queries on very large graphs, stabilizing the ingestion pipeline.
- Drove a security hardening effort covering cross-tenant Neo4j isolation, encrypted connections, and SSE-KMS encryption at rest.
Mactores Cognition Inc.
RemoteSoftware Engineer
Dec 2023 to Dec 2025- Delivered 8 projects across data platform and Generative AI engineering for enterprise clients.
- Upgraded the Apache APISIX gateway from v2.x to v3.1.0 with the APISIX Ingress Controller, etcd, and Prometheus and Grafana observability; validated graceful reloads by holding a 20 GB in-flight upload until the transfer completed, and benchmarked throughput with warp against nginx defaults.
- Built AWS data lake pipelines on Glue 4.0 (up to 20 concurrent G.2X jobs) with real-time ingestion through Kinesis Firehose into S3, and cut spend across Timestream, Glue, and CloudWatch.
- Built ML and GenAI pipelines for a lead-scoring classifier on SageMaker batch inference of ~100k/hr leads and a content generation service with tone control, orchestrated with Step Functions, Lambda, Glue 5.0, and RDS PostgreSQL.
- Designed Generative AI use cases for a telecom client (recommender, fraud detection, FAQ chatbot), and shipped the FAQ chatbot to production.
- Designed a dual-model system (GPT-OSS 20B with a Vision Transformer) for image classification, fine-tuned with QLoRA and OpenSearch embeddings. Benchmarked Nova Pro, Nova Micro, Nova Lite, and OSS GPT-20B and recommended Nova Pro on Trainium for accuracy and cost, targeting a 25 to 40 percent cut in manual workload with sub-second responses.
Software Engineer (Freelance)
Apr 2022 to Dec 2023- Delivered backend and cloud engineering for client projects, including an OAuth integration and demo environments deployed on AWS Amplify and EC2.
- Built and hardened API gateway infrastructure on Kubernetes with Apache APISIX, OpenAM, and Prometheus and Grafana monitoring.
Skills
Languages
- Go
- Python
- Lua
- C++
- Rust
- Zig
- Java
AI & Agents
- Strands SDK
- Claude SDK
- LangChain
- LangGraph
- Amazon Bedrock
- QLoRA fine-tuning
Cloud & Infra
- AWS
- EKS
- Terraform
- Temporal
- GitOps
Data & Search
- OpenSearch
- pgvector
- Neo4j
- Kafka / MSK
- AWS Glue
Platform & DevOps
- Kubernetes
- Apache APISIX
- GitHub
- Prometheus & Grafana
Credentials
Education. B.E. in Computer Science and Engineering, NIE Institute of Technology, Mysore, India, 2020 to 2024.
Certification. Claude Certified Architect, Foundations (Anthropic).
Let’s talk
I am always glad to talk about backend and AI systems architecture, trade notes on agentic systems, or think through a hard design problem. If any of that is useful to you, please get in touch.