Skip to content
View Saidheerajgollu's full-sized avatar

Highlights

  • Pro

Block or report Saidheerajgollu

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Saidheerajgollu/README.md

Hi, I'm Sai Dheeraj Gollu 👋

Software Engineer & AI Engineer · Open to full-time opportunities (F1 OPT)

Open to Work LinkedIn Portfolio Email


About Me

I'm a Software Engineer and AI Engineer based in San Jose, CA, currently on F1 OPT and actively looking for full-time Software Engineer / AI Engineer roles.

I'm finishing my M.S. in Software Engineering at San José State University (May 2026), and working full-time as a Software Engineer at Altrion.ai, building fintech backend infrastructure in Python/FastAPI and React/TypeScript. Before that, I spent two years building distributed Java microservices at Aviac Technologies.

My focus is backend and distributed systems, and applied AI — RAG pipelines, agentic tool-use harnesses, and the Claude/LangChain stack. I also contribute fixes to open-source projects on the side: four merged PRs into Andrew Ng's openworker, a merged accessibility fix in a 30k★ UI library, and a diagnosis + credited fix for a data-loss bug in a 14k★ agent harness.


Right Now

  • 💼 Software Engineer at Altrion.ai — multi-provider fintech data layer (Plaid, Coinbase, Stripe, Moralis DeFi), 200+ financial instruments, sub-200ms under concurrent load
  • 🎓 M.S. Software Engineering, San José State University — May 2026
  • 🛂 On F1 OPT, open to full-time Software Engineer / AI Engineer roles
  • ⚙️ Building TSDLite on the side — 1.1M points/sec, zero data loss
  • 🔍 Contributing fixes to production AI/OSS codebases (see below)

Stack

Languages Python · Java · C++ · TypeScript · JavaScript · Go · SQL
Backend & Systems FastAPI · Spring Boot · Node/Express · Apache Kafka · Concurrency & Multithreading · Unix/Linux
AI / NLP LangChain · Claude API · RAG (BM25 + embeddings + RRF) · HuggingFace · BGE-large · ModernBERT
Data PostgreSQL (composite indexing) · Redis · MongoDB · Neo4j · Elasticsearch
Cloud & DevOps AWS EC2 · GCP (Vertex AI, A100s) · Docker · Kubernetes · GitHub Actions · Jenkins
Frontend React 19 · TypeScript · Vite · Tailwind CSS · Recharts · Socket.IO

Featured Projects

TSDLite — large-scale time-series ingestion engine in C++20 Bounded ring buffer, concurrent data access, batched PostgreSQL writes. Sustains 1.1M points/sec wire acceptance with zero data loss under 8x burst overload, 141k rows/sec sustained DB throughput (35x faster than naive writes), 64 automated tests, full CI/CD, Docker/K8s.

Thrilochana — information retrieval & NLP platform, live on the App Store A 6-stage hybrid retrieval pipeline (BM25, embeddings, Reciprocal Rank Fusion) hitting 95%+ accuracy across 100+ document catalogs, plus a Plan-Execute-Reflect agentic harness spanning 54 tool integrations that cut processing time 97% against a ground-truth evaluation layer.

job-hunter-project — self-hosted job aggregator Polls ATS boards directly — Greenhouse, Lever, Ashby, Workday, SmartRecruiters, Workable, Recruitee — so matching roles arrive by push before they ever hit the big job boards. Next.js, Node/Express, Supabase Postgres.


Professional Experience

Software Engineer · Altrion — Jul 2026 – Present, San Jose, CA Full-stack fintech infrastructure in Python/FastAPI and React 19/TypeScript, shipped weekly to AWS EC2 via GitHub Actions. Designed a resilient integration layer across 10+ third-party APIs (Plaid, Coinbase, Stripe, Moralis DeFi) with per-step error isolation, so a single provider outage never fails the broader pipeline. Cursor-based incremental sync with per-source partial-failure handling — zero unplanned downtime across weekly releases.

Software Engineer · Aviac Technologies — Jan 2022 – Dec 2023, India Distributed Java/Spring Boot microservices; cut API response time 30% across 500+ daily requests with connection pooling and async processing. Diagnosed a silent Kafka message-drop defect via structured log analysis and fixed it with idempotent consumers and dead-letter routing, restoring 99.9% event delivery across 3 services. Took test coverage from 40% to 90%, ran 50+ code reviews, zero unplanned outages across 20+ Kubernetes releases.


Open Source Contributions

andrewyng/openworker ★17k · Andrew Ng's agent harness — 4 merged PRs, listed contributor

  • #55 rejected path-traversal session IDs in the conversation store
  • #56 made the conversation loader tolerate one corrupt line instead of losing the whole .jsonl
  • #379 fixed a race where a scheduled task could fire twice if an approval landed mid-tick
  • #24 fixed inbox reply-intent parsing to key off the leading word instead of substring matching

heroui-inc/heroui ★30k — #6741 merged, approved by a core maintainer Removed stray aria-labels left on aria-hidden decorative icons — an accessibility fix.

yc-software/qm ★14k agent harness — credited co-author on a merged fix Found that the Postgres-backed durable map JSON-stringifies into jsonb, which rejects NUL bytes and lone surrogate pairs — the exact kind of string the system's own truncation helpers manufacture when a clamp lands mid-emoji. Wrote the ADR describing the failure and the fix; the maintainer folded it into a merged batch PR crediting me as co-author.

In review

  • stanfordnlp/dspy #10301configure_cache() had no fallback when disk-cache init failed; added one
  • run-llama/llama_index #22857 — traced an nltk 3.10.3 hardlink guard that silently broke punkt_tab/RAKE keyword extraction, fixed it at the root instead of patching call sites

Certifications

🎓 Claude Certified Developer – Foundations, Anthropic (Aug 2026)


Always up for talking distributed systems, retrieval pipelines, or why your logs are lying to you.

Pinned Loading

  1. openworker openworker Public

    Forked from andrewyng/openworker

    Python