OPEN TO WORK · RELOCATION / REMOTE
CASE FILE · NLP-2026 YEREVAN · AM

Albert
Hakobyan

AI ENGINEER · AGENTIC SYSTEMS · NLP · BSDS '26

AI engineer building agentic systems, retrieval-augmented generation, and efficient models for low-resource languages. Capstone author of the first neural system for Armenian participle punctuation: I distill a frontier LLM's judgment into a 48.5M-parameter BiLSTM and mBERT ensemble that runs 1000+ sentences/sec on a laptop CPU, within 2.5% of the teacher, at zero marginal cost.

⚡ ENTER THE LAB → CV ⌁ TRANSMIT A SIGNAL
0
SKILLS MAPPED
9
KNOWLEDGE DOMAINS
1ST
NEURAL HY PUNCTUATION SYSTEM
🏆 
AUA ACSE RESEARCH POSTER SHOWCASE
WINNER · MAY 8

DOSSIER

[ ID. 001 · OPERATOR PROFILE ]
SUBJECT
Albert Hakobyan
B.S. Data Science
American University of Armenia · 2022–2026
Track: Business Analytics
SPECIALIZATION
Low-resource NLP.
Knowledge distillation · token classification · tokenizer engineering · transformer fine-tuning. Armenian language tooling.
COORDINATES
Yerevan, Armenia
albert_hakobyan2@edu.aua.am
Supervisor: Sachin Kumar
Linguist: Anahit Hovhannisyan

CASE FILE · NLP-2026

[ CAPSTONE · BSDS 2026 ]
PROTOCOL · DISTILLATION

KNOWLEDGE DISTILLATION FOR ARMENIAN PARTICIPLE
PHRASE PUNCTUATION RESTORATION

From LLM Teacher to Neural Student Models · Akian College of Science and Engineering, AUA
MACRO-F1
0.675
ENSEMBLE
VS TEACHER
−2.5%
p=0.11 N.S.
INFERENCE
1000+
SENT/SEC · CPU
TRAINING DATA
112K
ANNOT. PAIRS
▌ PIPELINE
OSCAR 23.01
22M sent.
→
AFFIX FILTER
4.45M cand.
→
STANZA POS
120K filtered
→
GEMINI 2.5 ★
CoT teacher
→
DISTILL
BiLSTM + mBERT
→
ENSEMBLE
α=0.45/0.55
▌ KEY FINDINGS
01
Ensemble matches teacher. Macro-F1 0.675 vs Gemini 0.700 (Δ=2.5%). McNemar p=0.11 — not statistically significant.
02
Student beats teacher on COMMA_AFTER. F1 = 0.495 vs 0.450. Student filtered out teacher noise.
03
Depth > language specificity. 12-layer multilingual mBERT (0.519) outperforms 6-layer Armenian-specific HyeBERT (0.326).
04
Zero inference cost. BiLSTM runs 1000+ sentences/sec on a laptop CPU. Teacher costs ≈ $0.003 per sentence.
05
Teacher-agnostic. Swapping to gemini-3-flash-preview (+5.3pts) requires zero architectural changes.
▌ THE PROBLEM

Armenian participle phrases ending in -ելով / -ալով / -ած require position-dependent punctuation. Errors change meaning.

Նա տեսնելով Արմենին տխրեց
↓
Նա, տեսնելով Արմենին, տխրեց:

▌ BENCHMARK · SHTEMARAN 292
BiLSTM
0.366
HyeBERT
0.326
mBERT
0.519
Ensemble ★
0.675
Gemini
0.700
▌ RULES R1–R5
POSITIONMARK
Intraposition, p.phrase ,
Pre-positionp.phrase ՝ V
Post-positionV ՝ p.phrase
Adverbialadv , R1
Relative, rp , R1

EXPERIMENT · TOKENIZER SURGERY

[ NLP GROUP PROJECT · QWEN2.5-0.5B ]

Grafted 30,766 Armenian tokens onto Qwen2.5-0.5B. Trained custom SentencePiece tokenizers, initialized new embeddings three different ways, then recovered the model with LoRA rank-16 on 500K Armenian lines. Final perplexity 8.33 · token count reduced 78.3%.

01
Analysis
TOKENIZER
FERTILITY
Benchmarked 9 tokenizers on 25,621 lines / 516,860 words. Spread 6.5×.
Best
2.18
Worst
14.26
02
Training
CUSTOM
TOKENIZER
Trained 6 SentencePiece variants on 5M sentences. All beat XLM-R.
Fertility
1.67
UNK
0%
03
Surgery
VOCAB
GRAFTING
Extended Qwen2.5 vocab 151k → 182k. 78.3% token reduction.
New tok.
30,766
Best PPL
24.4K
04
Recovery
LoRA
FINE-TUNE
Rank-16 adapters across 24 layers, 500K lines, cosine schedule. PPL 8.33.
Rank
16
Adapter
~9MB

FERTILITY · TOKENS / WORD

Lower is better. We measured how aggressively each tokenizer fragments Armenian text against 516,860 words from CC-100. Our trained SentencePiece BPE-32k beats every baseline by > 23%.

bpe_32k (ours) ★
1.67
XLM-R
2.18
mBERT
2.41
LLaMA-2
3.65
LLaMA-3
4.92
Qwen2.5 (base)
7.81
GPT-2
14.26
▌ QWEN2.5-0.5B — SURGERY REPORT
Base modelQwen2.5-0.5B
Parameters494M
Layers / Hidden24 / 896
AttentionGQA (14Q / 2KV)
Vocab in → out151,665 → 182,431
New tokens30,766
Init strategyHeuristic / FOCUS
LoRA targetQ · K · V · O — ×24
Train data500K Armenian lines
Final PPL (HY)8.33
▌ TEAM — Albert Hakobyan Levon Gevorgyan Robert Gadukyan Silva Vardanyan

ARCHIVE · THE VAULT

[ HYBRID RAG ENGINE · NOW OPEN SOURCE · FREE / LOCAL ]
IN SERVICE · DAILY DRIVER ◈ NOW OPEN SOURCE

HYBRID-RETRIEVAL RAG OVER
MY ENTIRE EDUCATION

Grounded, cited answers drawn from my own Obsidian vault, not the open internet. My private instance indexes a 100GB knowledge base of everything I have studied, with citations back to the exact page or heading. The engine behind it is now public: a config-driven hybrid-retrieval stack anyone can point at their own corpus. Zero running cost.
OPEN SOURCE Advanced-Obsidian-RAG PRIVATE INSTANCE 175K+ chunks · never shipped
CHUNKS INDEXED
0
4,500+ DOCS · MINE
VAULT SIZE
100GB
OBSIDIAN KNOWLEDGE BASE
KNOWLEDGE DOMAINS
9+
NLP → DEVOPS
RUNNING COST
$0
FREE / LOCAL STACK
▌ RETRIEVAL PIPELINE
HyDE
query expansion
→
DENSE
ChromaDB · top-20
→
SPARSE
BM25 · top-20
→
RRF FUSION
+ domain boost
→
RERANK
cross-encoder · top-5
→
GROUNDED GEN
[n] cites · confidence
▌ ENGINEERING LOG
01
Serve agents JSON, serve humans HTML. The moment an AI agent tried to "click" the web UI, the vault grew a warm JSON API — one HTTP call now replaces an entire browser session.
02
Memory-bounded indexing. The sparse index was rebuilt on sparse-matrix foundations, cutting its memory footprint several-fold with zero change in retrieval behavior.
03
Idempotent ingestion. Content-hashed document IDs make every ingest pass safely re-runnable — new material appends, nothing duplicates, and both indexes rebuild from one source of truth.
04
The cheapest lever wins. Large-scale corrections are applied as in-place metadata updates rather than re-embedding the corpus — maintenance costs minutes, not compute.
▌ THE MISSION

"What do I know about X? How did I implement Y?" The vault answers from my own materials — with citations back to the exact chapter, page, or heading.

Every answer carries its source.

▌ INSIDE THE VAULT

Lecture notes · 280+ textbooks · passed coursework · current-course slides · code notebooks · OCR'd scans · a self-study software-engineering library — every format parsed into one unified chunk schema.

▌ ONE ENGINE · MANY DOORS

CLI · a Corpus console for ingest, retag and live querying · a warm FastAPI JSON API for agents with schema discovery. Every retrieval knob is a config preset, and a 3-tier eval suite scores retrieval, answers, and confidence calibration.

▌ STACK — ChromaDB · bge-small-en-v1.5 (local CPU embeddings) · bm25s · RRF fusion · cross-encoder reranking · HyDE · Tesseract / DeepSeek-OCR · FastAPI · Docker Compose · versioned YAML prompts
Repo Advanced-Obsidian-RAG License LGPL-2.1 Deploy docker compose up Corpus 4,500+ docs · 17 code langs

AGENT · ARGUS

[ LANGGRAPH RESEARCH ENGINE · TELEGRAM · CITED REPORTS ]
PROTOCOL · DEEP RESEARCH

ONE TELEGRAM COMMAND IN —
A CITED RESEARCH REPORT OUT

A multi-agent research engine that turns a single question into a verified, cited report in Markdown and PDF. Live multi-source discovery, evidence-first synthesis, a three-judge review panel, and two human-in-the-loop gates — resumable across restarts, running on free-tier models.
GITHUB argus-research-bot
GRAPH NODES
10+
LANGGRAPH ENGINE
REVIEW PANEL
3
JUDGES · MERGE
HUMAN GATES
2
PLAN · REPORT
CODEBASE
21.5K
LOC · 39 TEST FILES
▌ RESEARCH GRAPH
INTAKE
quick / deep
→
SCOUT
multi-source
→
⟨ PLAN GATE ⟩
human approve
→
RESEARCH
fetch · digest
→
COMPOSE
parallel writers
→
3-JUDGE PANEL
revise loop
→
⟨ REPORT GATE ⟩
preview
→
MD + PDF
delivered
▌ SYSTEM DESIGN
01
Evidence-first synthesis. A cheap-tier model reads every fetched document into structured evidence notes — claims, supporting quotes, a 0–5 relevance score, a stance — and section writers compose only from those notes using [n] source IDs. Nothing reaches the report unread.
02
A plan gate you can trust. The graph pauses after live discovery, so the plan preview shows real sources search actually found. The brief never contains URLs, so there is nothing for the model to hallucinate.
03
Three judges, one verdict. Grounding, coverage, and precision judges score every draft with family-diverse model routing; a deterministic merge forces section-targeted revision on grounding failures, capped to stay bounded.
04
Discovery past text. Fetches and transcribes video from YouTube, X, Reddit and Instagram (async yt-dlp, captions or local faster-whisper ASR) and folds transcripts into the same evidence base as crawled articles.
05
Citation integrity + resumability. A 5-strategy URL verification cascade with domain-trust scoring guards every claim, and each run is checkpointed in SQLite (async) so it survives bot restarts.
▌ THE FLOW

Send one command. Argus scouts, plans, researches, writes, reviews, and delivers — pausing only twice to let you edit, approve the plan and preview the report before it ships. There are options to extend or revise the outputs. Media download feature enables an option to transcribe the media (if no tr. available -> via whisper), and to append the txt content to the research materials

▌ MODEL-TIER ROUTING

Three tiers — cheap · strong · judge — map onto whatever models the proxy exposes. Every call is logged as requested vs. served, with graceful fallback to auto. Swap the whole roster from config.

▌ COMMAND DECK
COMMANDACTION
/researchdeep cited report
/askquick answer
/statusrun progress
/fetch url..download links (media from: youtube/X/instagram/reddit)
/transcriptcaptions for /find picks
/qualitychanges the download quality
/cancelabort run
▌ STACK — Python · LangGraph · python-telegram-bot (async) · Exa · DDGS · arXiv · GitHub · crawl4ai · yt-dlp · faster-whisper · ReportLab · SQLite (AsyncSqliteSaver)
Repo argus-research-bot Scale ~21.5K LOC · 39 test files CI on every push Discovery Exa · DDGS · arXiv · GitHub

AGENT · TEACHING ASSISTANT

[ TELEGRAM BOT · AUA NLP PROJECT · LOCAL LLM · MIT ]
APPLIED · DEPLOYED

LECTURE SLIDES IN —
STUDY SESSION OUT

An agentic Telegram bot: upload a lecture PDF, receive a personalized, reviewer-approved deep-work study plan in your inbox · powered by a local Mistral-Nemo 12B — zero API cost, private by default · works on slides from any subject
▌ SIX-AGENT PIPELINE
SLIDE PARSER
PyMuPDF · summary
→
CONCEPT MAP
5–7 core concepts
→
WEB RESEARCH
DuckDuckGo · 3–5 links
→
PLANNER
timed session
→
REVIEWER
QA · pass / revise
→
EMAIL
SMTP · on approval
▌ CAPABILITIES
01
Human-in-the-loop control. Inline Send / Adjust / Apply-Fixes buttons: free-text feedback revises the plan, or the reviewer's own suggestions are auto-fed back to produce a corrected V2 — looping until the user is satisfied. Email is blocked until the QA verdict is PASS.
02
Self-reviewing agent. A dedicated reviewer agent audits every plan for timing, grounding in the slides, realism, and completeness before anything ships.
03
Concept-aware research. Web queries are derived from the LLM-built concept map, so external resources target the lecture's actual ideas — each link delivered with a justification.
04
Bilingual output. Session plans generated in English or Armenian, chosen during the /plan flow alongside duration, audience, and delivery email.
05
Local-first serving. Mistral-Nemo Instruct 12B (Q4_K_M GGUF · 32K context) behind an OpenAI-compatible KoboldCpp endpoint — swap models or servers by changing one env var.
▌ UTILITY

Turns any raw lecture deck into a ready-to-run deep-work session — objectives, timed exercises, and vetted external resources — in one Telegram conversation. The study plan a TA would write, generated and quality-checked on a laptop.

▌ COMMAND DECK
COMMANDACTION
/planfull pipeline run
/conceptmapmap slide concepts
/researchstandalone web search
/statuspipeline progress
/sendre-send approved plan
▌ STACK

Python · python-telegram-bot (async) · PyMuPDF · DuckDuckGo search · SMTP delivery · KoboldCpp / llama.cpp serving — MIT-licensed and fully reproducible.

CLASSIFIED · IN THE LAB

[ 3 FILES · DECLASSIFICATION PENDING ]
CLASS
IFIED
FILE · NLP-2026-01
IN PROGRESS██████
CLASS
IFIED
FILE · NLP-2026-02
IN PROGRESS████
CLASS
IFIED
FILE · NLP-2026-03
IN PROGRESS███████
● THREE PROJECTS UNDER ACTIVE DEVELOPMENT · DETAILS TO BE DECLASSIFIED SOON

DOMAINS

[ 684 SKILLS · 9 KNOWLEDGE FIELDS · 24 AUA COURSES ]
⌖
684 SKILLS · ALL DOMAINS

INSTRUMENTS

[ STACK · TOOLS · ENVIRONMENT ]
▌ LANGUAGES
Python · R · SQL · T-SQL · DAX
▌ ML / DL
PyTorch · TensorFlow · scikit-learn · Hugging Face · Gymnasium
▌ NLP
NLTK · spaCy · Transformers · mBERT · Trax · Stanza · SentencePiece
▌ DATA
Pandas · NumPy · dplyr · tidyr · Stanza
▌ VISUALIZATION
Matplotlib · Seaborn · ggplot2 · Plotly · Power BI · Tableau
▌ DATABASES
PostgreSQL · MySQL · MongoDB · SQL Server · SQLAlchemy
▌ BACKEND
FastAPI · Pydantic · Docker · Docker Compose
▌ DEV TOOLS
Git · GitHub · Obsidian · Streamlit · MkDocs
▌ APPLIED
Open-source RAG engine (hybrid retrieval · 175K+ chunks) · Agentic Telegram bots (LangGraph · local LLM) · Web scraping · API integration
▌ EDUCATION
AUA

B.S. IN DATA SCIENCE

AMERICAN UNIVERSITY OF ARMENIA · 2022 — 2026 · TRACK: BUSINESS ANALYTICS

24 AUA courses spanning statistics, ML/AI, NLP, RL, time series, BI, marketing analytics, databases, visualization, and mathematical foundations. Capstone research in low-resource NLP.

▌ CERTIFICATIONS
VIEW ORIGINAL PDF →
TRANSMIT A SIGNAL.
REPLY TIME · YEREVAN HOURS · LAB ONLINE
EMAIL THE LAB LINKEDIN GITHUB HUGGING FACE CV