I bridge the gap between elite algorithmic problem-solving and production systems engineering[cite: 2]. My core focus lies at the intersection of autonomous AI execution layers, high-throughput RAG infrastructure, and performance-tuned backend systems[cite: 2].
I am looking for roles around AI Engineering, AI Infrastructure Engineering, AI Software Engineering, and High-Performance Backend Engineering for AI Products[cite: 2].
| Signal | Proof |
|---|---|
| Agentic Systems | Dual-agent parent-child orchestration (AlgoSentinel), LangGraph state routing, multi-agent supply-chain monitoring workflows. |
| Retrieval & Grounding | Hybrid token/vector engines, z-score fusion retrieval, cross-encoder reranking, and deterministic citation validation[cite: 2]. |
| Systems & Backend Depth | Custom Kubernetes CRD controllers (Go), edge-native middleware pipelines (TS), async event loops, and optimized PostgreSQL routing[cite: 2]. |
| Competitive Programming | CodeChef #6 India / #76 Global, LeetCode Guardian (Top 0.9% globally), Codeforces Grandmaster[cite: 2]. |
| Open Source | 35 authored PRs across 22 repositories (14 merged) resolving parser grammar, resource safety errors, and concurrency panics. |
| Project | Why it matters |
|---|---|
| AlgoSentinel | Autonomous multi-agent PR review daemon. Profiles changed code functions inside ephemeral Docker sandboxes to map growth curves and catch algorithmic complexity regressions ($O(n) \rightarrow O(n^2)$). |
| EdgeSentinel | Edge-native bot detection engine deployed directly to Cloudflare Workers with sub-10ms overhead. Deploys a 5-tier request classification cascade backed by Durable Objects rate-limiting to eliminate KV state race conditions. |
| Grounded Support Triage | Multi-domain LangGraph orchestration pipeline featuring hybrid vector/BM25 search, unified z-score fusion, and a programmatic verification layer checking hard facts to completely eliminate model hallucinations[cite: 2]. |
| TurboQuant KV Cache | PyTorch research implementation reproducing online vector quantization paper. Achieved a 3.3× KV cache compression ratio on GPT-2 Medium with only a +6.1% change in perplexity performance metrics. |
| MindSafe | Privacy-first clinical mental health screening dApp. Runs adaptive inference entirely client-side with zero data persistence, minting cryptographic commitment hashes on-chain via Midnight Network zero-knowledge (ZK) proofs. |
| sandbox-warm-pool-controller | Kubernetes custom CRD controller built in Go that implements a 6-state pod lifecycle pool to pre-warm hardware-isolated runtimes (gVisor/Kata), completely eliminating 1–5s container cold-start delays. |
Secondary Implementations: Avia Search, FounderOS, Nexus Feed, PinPoint AI, BTIFIT Platform, AI Traffic Shaper.
| Repository | PR | Area | Impact |
|---|---|---|---|
| mem0ai/mem0 | #3549 | AI Data Infrastructure | Engineered and integrated the native Azure AI Search vector database connector layer within the TypeScript SDK core (~646 LOC contribution). |
| sqlfluff/sqlfluff | #7163 | Parser Engineering | Extended the core PostgreSQL dialect grammar engine rules to properly parse explicit programmatic OPERATOR(schema.op) syntax overrides. |
| pandas-dev/pandas | #62592 | Core Test Engineering | Developed robust regression test suites safeguarding PyArrow datetime data-type data merges against unexpected duplicate-key drops. |
| interviewstreet/hiring-agent | #133 | Performance & Safety | Eliminated critical background memory leak vectors by converting raw PyMuPDF file object tracking loops into safe context-managed lifecycles[cite: 2]. |
| interviewstreet/hiring-agent | #129 | System Robustness | Enforced explicit UTF-8 encoding patterns on backend cache file write operations to guarantee Windows file system Unicode execution safety[cite: 2]. |
- Telemetry Streaming (
kubeflow/sdk#436): RefactoringLocalJob.logs(follow=True)execution loops into native generator streams to remove hidden stdout-blocking side effects and enable real-time training tracking telemetry. - Symbolic Mathematics (
sympy/sympy#29605): Patchingrefine_sign()logic architectures to fix assumption handling constraints for unallocated plain variables. - Test Correctness (
modelpack/model-spec#211): Resolving silent test passes on malformed JSON configuration examples by fixing error variable conditional evaluations.