Most engineers build
agents that forget.
I build the memory engines
that make them remember.
Hi, I'm Sai Likhith — Senior Software Engineer embedded in the AI Data Prep ecosystem at Airbnb. I build high-throughput multi-session GenAI orchestration drivers, active-learning labeling backends, and PII-sanitized episodic memory systems. 8+ years experience bridging deep retrieval mechanics with production-scale data pipelines.
7 years running
SAI-AGENTIC-MEMORY-SPEC
Node Performance Telemetry
Pipeline Architectures & Workloads
Production workloads engineered, optimized, and deployed to run semantic retrieval, active-learning, and synthetic data generation at scale.
BPI-VIRTUAL-ANALYST
Multi-session LLM state router and evaluation sandbox. Converts high-dimensional raw case logs into PII-sanitized episodic memory traces, routing across 30+ LLMs using schema-enforced JSON validation.
- JWT KeepAlive TTL monitors to prevent token expirations during multi-hour active-agent evaluation sweeps.
- Microsoft Presidio semantic filtration pipeline sanitizing 12 entity types to ensure PII-safe vector embeddings.
- Two-layer JSON schema alignment and validation system resolving mid-stream LLM generation truncations.
REDPEN-LABELING-INFRA
Active-learning annotation platform and synthetic dataset generator. Orchestrates dataset ingestion, model-assisted labeling (MAL) benchmarks, and hourly evaluation exports.
- Designed a 5-layer hourly/daily delta DAG system with high-precision activity checks for label export syncing.
- Hardened client base wrapper using custom exponential backoff retry-after decorators.
- Pydantic state models preventing data truncation on 18+ digit identifiers by forcing string types.
LILLY-DMS-PORTAL
Procedural memory audit logging system. Engineered database triggers capturing document modifications as JSON diffs to guarantee compliance workflows and data lifecycle audits.
- Implemented database-level master-data audit triggers capturing JSON diffs directly.
- Configured idempotent Flyway migrations with schema checks to prevent deployment blocks.
- CI/CD workflows deploying containerized application instances across QA/Prod namespaces.
SHELL-NLP-PIPELINE
Distributed ETL pipelines and semantic retrieval classification algorithms. Deployed custom text-classification NLP models on high-scale SageMaker endpoints.
- Published SPE ATCE Conference research paper (22ATCE-P-663-SPE) on ML reusable components.
- NLP text classification engine using rule-based regex parsing + tokenization.
- Optimized Databricks and PySpark ETL query caches to clear 200+ query-planning bottlenecks.
SWA-CLUSTER-MONITOR
Telemetry reporting and cluster monitoring for predictive analytics pipelines deployed on multi-pod Kubernetes clusters.
- Dockerized SageMaker & S3 storage hooks for multi-pod Kubernetes scheduling.
- Developed high-throughput indexers syncing structured aircraft records into Elasticsearch.
- Built telemetry hooks capturing container crashes and forwarding traces to Datadog.
Peer System Verifications
Documented testimonies and SLA approvals from platform leads and deployment partners confirming architectural competence.
AUDIT::AMEET-SHINDE
“Sai has been an outstanding partner in the deployment of the BPIVA tool, and I want to take a moment to recognize his incredible contributions. Thanks to Sai's efforts, the BPIVA tool has had a significant impact on reducing non-value-added work, enabling the BPI team to shift their focus to high-impact, actionable tasks exactly where their energy should be. What truly sets Sai apart is his deep understanding of technology combined with his ability to quickly grasp tool requirements and translate them into real solutions. He doesn't just deliver, he continuously looks for ways to enhance and upgrade the tool's capabilities, ensuring it evolves alongside our team's needs.”
AUDIT::JEREMY-CHUA
“A huge shoutout to Sai for going above and beyond in supporting our new HALO [Human Annotation] team in AirCover! 🙌 From answering my Labelbox questions to proactively flagging solutions I hadn't even thought to ask about — Sai made the whole process so much smoother. This support has been instrumental in helping our team in AirCover get off the ground and hit the ground running. Really appreciate you, Sai!”
AUDIT::LORI-BARBER
“Thank you, Sai, for being invaluable to setting up the Luxe labelbox project and working so quickly to resolve matters. I look forward to working more closely with you in the coming months.”
AUDIT::ALEJANDRO-VIRRUETA
“Thanks for covering the on-call today! Good job investigating the first ticket!”
FTE Node Compatibility Spec
Audit mapping Sai Likhith Kanuparthi's production credentials against requirements for NVIDIA's Agentic Memory Engineering team.
Agent Memory System Fit
With 8+ years of production experience spanning Airbnb's GenAI platform and Fortune 50 enterprises, Sai Likhith is positioned to immediately contribute to NVIDIA's agentic memory and synthetic evaluation frameworks.
Allocate Compute Session
Transmit server configuration requirements or scheduling requests. Packets are parsed and routed directly to Sai's active terminal.
Transmitted payload is stored with TLS encryption and forwards details directly to sailikhithcse@gmail.com.